|
Kernel Traffic Latest | Archives | People | Topics |
Wine Latest | Archives | People | Topics |
GNUe Latest | Archives | People | Topics |
| Czech |
| Home | News | RSS Feeds | Mailing Lists | Authors Info | Mirrors | Stalled Traffic |
Halloween is /nick 'costume night' in #gnuenterprise - "<RMS> software should be Free - tell your friends ;)" - "<LarryEllison> now, how can I maintain my mansion - giving away my software - that's my INTELLECTUAL PROPERTY boy!"
Table Of Contents
Introduction
This Cousin covers the three main mailing lists for the GNU Enterprise project, plus the the #gnuenterprise IRC channel.
1. Configuring Oracle to listen to GNUe
30 Oct 2002 Archive Link: "[IRC] 31 Oct 2002"
Topics: Common
People: Jason Cater
Further to Issue #53, Section #2 (24 Oct 2002: GNUe support for Oracle 9i) , it was noted that GNUe looked for a particular Oracle listener, using the service name from connections.conf. Oracle's own SQL*Plus tools looked for the default listener. This was with running both SQL*Plus and GNUe on the same machine as the database server. Jason Cater (jcater) explained "when running on the server, sqlplus can use interprocess communications to talk with the server - we can't - we can only do it via tcp/ip" . This had never been an issue for him, as "I am running Oracle in client/server - nothing but oracle runs on the server" . This meant that the connections information was taken from the TNSNAMES.ORA file as well as listener.ora. There should be no problem running GNUe and the Oracle database on the same machine, but "you'll still need a listener running though : :)" . He volunteered to send his Oracle configuration files to help. The problem turned out to be that the name of the service was different to the (SERVICE=) entry in the tnsnames.ora file, which was concerning, as this file had been automatically generated by the Oracle tools, so therefore should not have gone wrong. It was asked if there was any way to verify the file. Jason said "I don't know - I had a hard time setting mine up the first time - when I finally got a working version several years ago - I never touched it again" .
2. New CVS on savannah for GNUe Small Business and RentFree
30 Oct 2002 - 2 Nov 2002 Archive Link: "[IRC] 31 Oct 2002"
Topics: Small Business, RentFree, Designer
People: Jason Cater, Andrew Mitchell, Derek Neighbors, Charles Rouzer, Nicholas Lee
Further to Issue #53, Section #26 (27 Oct 2002: GNUe small business packages) , Jason Cater (jcater) said that savannah hackers had queried his project submissions, asking for an "URL to the source code tarball" . Andrew Mitchell (ajmitch) said this was "just to check licensing" . Jason said "I am creating new projects, though - how can they check licensing for something I'm going to do?" . Andrew said "you said for RentFree that you had source, they'll check that the files have the proper copyright into" . Jason said "I already indicated that it was gpl" . Andrew replied "as do a lot of people - but people seem to have a knack for missing copyright headers on source files. Projects that i've submitted have been rejected & told to resubmit, too. Do you have any code/forms/stuff for gnue-sb?" If not, he would need to explain that to them. Derek Neighbors (revDeke) noted that GNUe seemed to be the only GNU project that was consciencious in getting copyright assignments from contributors, so it was a bit galling that "they are worried one of our SUB PROJECTS might not have good license?" Jason said he did not mind so much that they needed clarification on something, but was surprised that it had taken two days to get back to him on it. Andrew said "if you saw the pile of projects that are in the submission queue..." .
The next day, Andrew Mitchell (ajmitch) noted that Jason's submissions were being discussed on the savannah-hackers mailing list. He reminded Jason that he would have "to resubmit, which means using the URL provided - this is because once they have been declined for whatever reason, the system is designed for you to have to resubmit :)" Jason was not very impressed with this - "no history information? bah" . Andrew said that savannah had been forked from the last public version of sourceforge, so shared many of its limitations. Jason said the free software way was to "hack it to be better" .
Jason was also confused, as "They asked me if I wanted to be a part of the GNU project on the registration screen" , which he had replied yes to. They were now asking for proof of this, but the question had been whether he wanted to be, not whether he already was. Andrew said "i told them that GNUe subprojects should be considered part of GNUe (as if my opinion matters ;)" .
Jason also raised the issue of how he should mark the code as being under the GPL (GNU General Public License) - "I have a COPYING file - however - All the "code" is really GFD files - is it going to come back to bite me that there's not copyright headers in the GFD files? or is the COPYING file enough" ? Andrew Mitchell (ajmitch) said "the COPYING file states that you should have headers in each source" code file. Jason said the "thing is, the GFD files are automatically generated" by Designer - "so keeping a header in there will be a biotch" . Andrew suggested "perhaps go the nasty way & have a license tag for forms - so that designer can generate copyright header in xml comments" .
Derek Neighbors (derek) said he thought that "the gfds should have the copyright header - its not MANDATORY - the COPYING file is enough - but it is better to have in GFDs" as well. Jason said he would put them in manually for now. Derek said the long-term solution was "in designer allow a 'preferences' that lets you put in custom headers - and wehn you make gfd's you can choose from those" . Jason said "I think once I get "projects" working in Designer this will be the appropriate place for that - as on the "project" config screen, you could say "place xxxx copyright header in all G?D files"" . Derek said this "is GREAT for two reasons - a. you dont have to manually fight designer - b. it will add headers so you dont forget :) - /me is VERY guilty of b" . Jason felt it should not be "a system-wide preference - as I work on all kinds of projects w/gnue - public, gpl'd projects - internal projects" and so on. Derek said "i would do a 'profiles' thing and then be able to set it by project - i.e. you could have a GPL profile, BSD, company foo etc" .
The next day, Derek noted that the project page for GNUe Small Business was now on savannah - "i have code to actually already check into cvs - just getting cvs ready - hopefully by monday it will be 'official' - and you can checkout code and everything for contacts and product management" . He explained "its a SUB project of GNUe - the 'official' gnue applications will be made using the appserver and will be much more intensive - the 'small business' applications are 'official', but more a side addition - they will be released much sooner but will not use appserver" .
Derek also noted "we are starting the legal battles to get fedex and ups api for shipping management as free software - but regardless of how we release it that is being worked on as well - sales order, invoicing, shipping management, contact management, inventory/product management should be within next 30 - 60 days" . Charles Rouzer (Mr_You) asked "how are other free solutions connecting to UPS/Fex?" Derek replied "they arent or they are doing so illegally - the way it looks right now (at least for fedex) is that if you write a solution you can not distribute it - whether its 'free software' or 'propreitary' software" . He said "we have some issues to resolve legally - we are going by the letter of the fedex contract (not us, but real lawyers reviewing) - we will likely try to communicate with them to get written consent to develop free software using the api's in some form or another - as the other alternative is to 'reverse engineer' and play samba like api madness - which we certainly could do if necessary" . He was not tempted to just "download the api and write something and distribute under gpl" as "if fedex found out and got pissed off, they could kill the project dead" . The worst case scenario was that the shipping management would have to be done as proprietary software "that interfaces with gnue" but "we would MUCH rather it be free software" .
Derek said that the 'code' for GNUe Small Business was "pretty much all schema's and xml files - which is what is great - entire applications beeing xml and db schema. Actually i think in the product management stuff for right now i only have 1 trigger :) - and that is to over come an oddity in my ui refreshing :)" to refresh a list of foreign keys for a dropdown, as discussed in Issue #53, Section #11 (24 Oct 2002: Updating drop-down lists and changeable runtime parameters) .
The next day, Derek cut and pasted his initial CVS check-in for GNUe Small Business. Nicholas Lee (esands) noted the GNUe Schema Definition (.gsd) files in the check-in and asked "I assume designer reads gsd files and outputs sql?" Derek said "kind of" - "gsd is an xml file - currently you use xslt and an xsl (style sheet) - and it makes a sql file." At the moment, he did this from the command line, but it was intended that Designer should do this directly in the future, and it might do so already. He noted the code for the XSLT was "all in gnue-common iirc" - "gnue/common/utils/xml2sql/ i think has the stylesheets" . Nicholas asked "what deb is PySablot?" . Derek said that the pysablot "debs arent in sid yet :( - working on it - they are on our website though :)" . He warned that the .gsd-handling code in CVS was temporarily broken as of time of writing, which Nicholas confirmed - Derek was using a slightly older copy that still worked, as "i have real world deadlines and others have needs that are watiing on us - so expect gnue-sb to move at fairly steady pace compared to how gnue apps have gone in the past" .
3. Application Server development
2 Nov 2002 Archive Link: "[IRC] 03 Nov 2002"
Topics: Application Server, Designer
People: Neil Tiffin, Reinhard Müller, Daniel Baumann
Neil Tiffin (neilt) said he wanted "to build business objects - and as long as we have a agreed aproach that works" , he was "happy to see the work progressing" - . Reinhard noted "we have finally agreed on storing the class definitions in the database - or to be more precise in "system business objects"" . Daniel Baumann (chillywilly) noted "well I have a parser commited for ODL that uses PLY in case no one noticed, but I need to make it parse into something like perhaps an AST" . Reinhard confirmed this had been discussed, "as well as about comparable things that only exist more or less as an idea - and we agreed that it is a good thing to have parsers for other formats - that can read "external" defined objects and "merge" that into the "main" repository as a separate command line tool" . Neil felt "the back end for storing business objects should be plugable just in case we want to change it" . Reinhard agreed - "that's why made it a separate "module"" .
Daniel said "I figured using an xml markup as an exchange format between designer and appserver would be nice too - since designer probably wants xml" . He pointed out that "it is supposed to be our IDE isn't it? jcater made it so you could write plugins" . Reinhard suggested "we could even have some "normal" form that accesses these metaobjects" .
Neil asked "can business objects be aggregates of other business objects?" - for example, "so the GUI can just request the sale order and get all the header and items on the sales order" at once. Reinhard suggested this might not be a good idea, as "the sales order might have 1000 items on it - and the gui doesn't want to load all 1000 at once" . Neil said he was used to business objects corresponding with a form - "this business object can be a real object or a proxy object for several real business objects - but this means that some of the complexity of the GUI is moved to the business object" . Reinhard felt "that the appserver shouldn't deal with visualisation of data" - "if we define those "form-bound" business objects - we move parts of the form definitions into appserver" . Neil agreed about visualisation, but said "if we dont" have compound business objects "we move linking all of the data to the form" - if links "also have to be defined by the form so the work to maintain it will be double" . "The forms or appserver will be fragile to breaking if one is changed but not the other" .
Reinhard felt "we will always have this issue as long as form definitions (XML) and object definitions (database) are maintained seperately - that a form definition doesn't match the underlying object" . Neil said that compound business objects should actually help with this, as the compound object could be changed to pick up a property from a different child business object, but still present it as the same field to the form. Reinhard understood this, but did not see "how this could work with 1:n instead of 1:1" . Neil admitted "thats is the downside - in that you hide the relationships in the proxy objects and the forms just deal with data - but the upside is that i can add a complex stored procedures to retreive data into and object and the form does not need to know the relationship - only that it should display the data" . Reinhard asked "but the form has to know that it's master detail doesn't it?" Otherwise "how does the form know that it can insert a new item into the order?" Neil said "the list is either read only or read write - if the list is added to then the business object maintains referrential integrity when it updates the db" .
Reinhard asked "how does the form know that i can add another item to the order - but i can not add another customer? well i can _change_ the customer" . Daniel Baumann (chillywilly) said "it shouldn't display stuff to add a customer then ;) - or the gui should be 'disabled' - forms has read-only widgets" . Reinhard suggested "what you are talking about means that we have some "array" type of property" . Daniel said "which in ODL would be a ref to a collection ;) - cause relationships and attributes are the object "properties"" .
Neil emphasised that these were just his initial thoughts "and i dont want to get in the way of progress" . Reinhard agreed - one of the themes of the meeting had been to reduce discussion in future and implement possible solutions - "look where we will have the _real_ problems - and improve that" . He felt that the discussion he and Jan had had about the AppServer API in Issue #53, Section #14 (26 Oct 2002: Performance and Overhead issues with AppServer) was a good example of this - "with jforms siesel found out we have to change it to make performance better - so we changed it - but if we hadn't done that prototype implementation - we would never ever have seen the real problem" .
4. Unusual error message from GDataObject
31 Oct 2002 - 4 Nov 2002 Archive Link: "[IRC] 01 Nov 2002"
Topics: Common
People: John Lenton, Andrew Mitchell, Jason Cater, James Thompson
John Lenton (Chipaca) reported an unusual error message - ""You've got your bar in my foo! And you've got your foo on my bar! Two great reams that ream well together!"" - "it's in __singleQuery - I'm trying to massage it into being more helpful - but I thought I'd check in here first; I can't be the first guy to come accross this" . Andrew Mitchell (ajmitch) suggested adding some debug statements "to print the traceback as well" John said "the error is "no results to fetch"" . Later, Jason Cater (jcater) said "if he's getting that message, then he's calling a virtual function that should be over-ridden by a subclass - /me fears they are messing around with GDataObject internals that they shouldn't be - kind of like me and electricity :)" .
The next day, John mentioned the error messaga again. James Thompson (jamest) said "i remember sticking in an error like that - but only when things failed rather hideously" . This had been removed from recent CVS versions. He related a (possibly urban legend) story about "a coder that left in a message to himself along the lines of "hey dumbass, you shouldn't be able to do that" - (this was a comercial product) - anyway it shipped with the error message accidentaly left in - some lady hit it :) Can you imagine the tech support call - "Yeah. Your program called me a dumbass." - "Well....is it true?"" .
Later, John reported that the error "seems to ocurr even with some of the samples from cvs (this is with a current gnue from your cvs)" . Jason asked "how are you triggering that message? I've never seen it. is it a specific function you are calling that triggers it - or a certain trigger" ? John said it was just "when accessing the db - it's in pgsql's extension triggers (or something like that" ). James said this was an "old error of mine - meant really bad thing happened that should never have been triggered - it's not in our cvs anymore IIRC" . John said he would check further. Jason said "something y'all have done is triggering it - I've never seen it triggered before" Unlike James, he though the error message was still in the CVS code, "but like you said, it's in a place that is never triggered - if you can reproduce in one of the samples, let us know - I've just never seen it - doesn't mean it doesn't happen :)"
The next day, John reported "I couldn't reproduce the reaming error on a clean gnue pulled last night from cvs with any of the samples in the forms/samples directory - so you've fixed it between when we pulled your cvs and last night, or we broke it in the same period (methinks the former)" .
5. Getting official Debian packages for pysablot
31 Oct 2002 Archive Link: "[IRC] 01 Nov 2002"
Topics: Reports
People: Derek Neighbors
Further to Issue #53, Section #30 (28 Oct 2002: Debian packages for GNUe into sid (unstable)) , Derek Neighbors (derek) said "i finally sent files upstream to pysablot guy and asked him to apply and make new tarball - hopefully he complies and you are able to make official debs real soon now (tm)" for the Debian unstable distribution (sid).
6. Copyright Assignments for GNUe
31 Oct 2002 Archive Link: "[IRC] 01 Nov 2002"
People: Derek Neighbors, Andrew Mitchell
Derek Neighbors (derek) needed to check copyright assignments for GNUe on one of the FSF servers. Andrew Mitchell (ajmitch) restored his account, setting up an ssh key for access, discussing some of the principles of key signing. Derek logged in, and said "we have a ton of assignments - i think we are half the copyright file for" the whole of the Free Software Foundation (FSF).
7. Triggers and cross tabs in Reports
31 Oct 2002 Archive Link: "[IRC] 01 Nov 2002"
Topics: Reports
People: Jason Cater, James Thompson
Jason Cater (jcater) said that GNUe Reports now supported triggers, thanks to James Thompson (jamest) - "0.1.0, here we come! /me isn't gonna guarantee just *how* well they are supported, but the basics are there" . Also, "I've started on a pivot table/cross tabulation example in report's samples/ directory - per the mailing list thread - I have a couple of bugs to get out" .
8. Symbolic link bug in CVS
31 Oct 2002 Archive Link: "[IRC] 01 Nov 2002"
Topics: Application Server
People: Dmitry Sorokin, Jason Cater, Andrew Mitchell, Jan Ischebeck, Derek Neighbors, ra3vat
Dmitry Sorokin (ra3vat) asked "what is the cause of cvs [checkout aborted]: reading gnue/appserver/grpc/GEAS.grpc: Too many levels of symbolic links - it was discussed earlier" . Jason Cater said "I have no idea what that's from - I removed that file from my directory when it happened to me - and updated again - but I have no idea what's causing it" . Andrew Mitchell (ajmitch) said "derek & i also got that error" . Derek Neighbors (derek) said he had asked savannah-hackers if they knew what the problem was in Issue #52, Section #11 (19 Oct 2002: Problems updating GNUe CVS server) , but they had not been able to help. Derek also said that Jan Ischebeck (siesel) had told him he knew what the problem was, but he was not sure whether Jan had actually fixed it yet. The issue was that ""GEAS.grpc" was replaced by a symbolic link pointing to itself" , causing an infinite loop. "The actual CVS version of setup-cvs.py now provides both an upgrade path AND removes the recursive symbolic link "GEAS.grpc"."
9. Re-using code in multiple triggers in a form
31 Oct 2002 - 3 Nov 2002 Archive Link: "[IRC] 01 Nov 2002"
Topics: Forms
People: John Lenton, James Thompson, Jason Cater
John Lenton (Chipaca) said he was trying to shift some code from a trigger to a function - despite the function containing the same code as the trigger, he could not get this to work. "I'm splitting the triggers into functions in a separate library because we're using the same old triggers again and again and again" . James Thompson (jamest) asked "have you tried using our import libraries instead to store the trigger? - .gfd files support trigger libraries - would that give you what you require ?" . John said "it'd make it much easier - but I don't think it'd make this fn work" . James said he would look at John's code - "as I did lots of trigger crap last night" . John said he was using a version of Forms that was older, as project papo's CVS was still a few weeks behind GNUe's CVS.
Later, Jason Cater (jcater) asked "stops working in what way?" John said it just "silently doesn't do anything" . Later, he asked if there was any documentation for the form libraries James had mentioned - "what I understood was that there was a general way of adding functionality to forms without having to import the library in every trigger, for example - maybe I understood too much" . Jason said "you can globally import files once on an On-Startup trigger - then it's available to every trigger in the form" , giving an example - this "works for variables too" . This was in the latest CVS version - "it's relatively new" .
John got an up-to-date CVS copy, and reported that Forms now generated a segmentation fault. James asked whether this was "an on-startup in GFTrigger.py?" He explained "GFTrigger.py died last nite in the cvs tree - all trigger code is in common now" .
Some days later, James said the way that trigger libraries worked by creating "a trigger lib file" containing the code for the triggers, then in the normal GNUe Forms Definition (.gfd) file, you would use an import-trigger tab, with attributes library="filename" and name="triggername". You could then refer to the trigger as if it had been defined locally. He was "working on the library system now" - "/me is altering the parser system to allow other gnue apps to do imports - the current cvs is forms only wrt imports - i should have any gnue apps able to do them" soon.
10. Kernel Cousins GNUe to go group-authored?
31 Oct 2002 Archive Link: "[IRC] 01 Nov 2002"
People: Peter Sullivan, Jason Cater
Peter Sullivan (psu) noted that the first anniversary issue of Kernel Cousins GNUe "has more words than issue 1 had bytes..." Jason Cater (jcater) said "good job - pretty soon, you might have to learn to delegate :)" Peter said that "group authorship is a distinct possibility" . He felt that Kernel Cousins "might be a good way for non-coders to get involved - without having to dig into the deeper recesses of the dcoumentation or the website (our other two big non-coding TODOs). Main problem is how I "tweak" the mechanisms of group-authoring a KC to fit IRC as the way the other lists work is that people claim "threads" via a central mailing list once they're finished - IRC != threads" He would "not want people to divert coding time to write KCs - as that rather misses the point" .
11. Multi-currency in General Ledger
31 Oct 2002 - 1 Nov 2002 Archive Link: "[IRC] 01 Nov 2002"
Topics: Financials (Accounting)
People: Peter Sullivan, Nick Rusnov, Jason Cater, Nicholas Lee, Reinhard Müller
Nicholas Lee (esands) asked about multi-currency General Ledger functionality. Peter Sullivan (psu) said "you decide on a base currency - then wherever you have an "amount" field normally - you have 3 - base amount, foreign amount, and foreign currency. Foreign currency AP & AR are the real funsters - as you have to hack gain/loss on settlement of open transactions. e.g. raise AR invoice for USD 100 ( = GBP 64) - eventually you get paid USD 100, but by then it is GBP 63 - so you have to write off the missing GBP 1 as a forex loss" . Nick Rusnov (nickr) suggested "see, if we had One World Currency we wouldn't have this problem" . Jason Cater (jcater) suggested "I vote on donuts - of course, currency that can spoil might make saving interesting :)" .
The next day, Nicholas said "For reconcilations I assume most multi-currency GLs keep a seperate ledger of (rate, amount) figures. What's the general practice for pricing a currency account on a given balance date? given that the base currency figure in that account represents and average of the prior exchanges, but the rate to realise the monies on a balance rate might result in a fx loss or gain." Peter, said "To me, the key principle is that you define a base currency - and then all transactions in GL store amount in base currency, amount in foreign currency, name of foreign currency. To get "true" trial balance you do select sum (base_amt) from balances ;" Nicholas said "That's need for tax purposes and need for in order to rationalise transactions within the GL. From a business management point of view though you need reconcile transactions and deal with making profit over multiple currencies" Peter felt "only if that's useful management information to you - I could imagine a biz not caring how much profit they made in Aussie dollars, just how much they made on product category X" .
Nicholas asked "What's the general rule of thumb in the UK regarding balance date base currency valutions of a foreign bank account?" Peter disclaimed "I'm not an expert on forex, but I believe this is where unrealised gain/loss on forex comes in" - "for balance sheet purposes you convert at prevailing exchnage rate at 31st Decembefr - The difference between this and what the amount in your books is constitues (IIRC) the unrelised gain or loss" . He gave a simple example of an unrealised gain. The "main issue (i am not a lawyer, especially a tax lawyer) is how taxman wants you to treat this" .
Nicholas asked what exchange rate to use - "the given rate on that day? or by the internal rate of sumif(GBP)/sumif(USD)" ? Peter said "I would always go far a "daily" rate in this day and age - with computers to do the work for us - monthly rates are less accurate - taxman might have a view too" . Nicholas said "I'll have to think how to use that in the context of hedging your transactions." Peter said "hedging if used properly is just a way of realising the potential gain/loss - i.e. take a forward position for when you expect the transaction to settle (for an invoice, typically 30 days)" . Nicholas said "Depends on the terms, for instance many of our transactions are 90 day - foreign exchanges ones that is" .
Nicholas asked "I assume a form of that triple is used" in the Accounts Payable and Accounts Receivable ledgers as well. Peter agreed - "once the transaction is settled (i.e. paid) you'd have a trigger to work out the *realised* gain/loss and post that right away" . Nicholas said "Actually the commerical system we use, generates the variances both on posting (month end) and payment" .
Reinhard Müller (reinhard) noted that, in Austria and Germany, "for the balance, you may not consider unrealized forex gains - and you _have_to_ consider unrealized forex losses - so you always have to use the "worst case" in the balance" . Nicholas was surprised "they don't try get more tax by requiring realisation of gains though. Must be nicer now you have the euro." Reinhard said the aim was to ensure "that the balance shouldn't look better than it really should be" . Nicholas said "Doesn't that depend on your point of view? Although the balance in both currencies in the balance sheet is an average of past transactions. At given moment it can be realised into the base currency. All past rates are in the past and all future rates unknown. Thus there is no true better or worse. Still I agree its better just realising fx losses. Otherwise it could cause undue hardship on a businesses cashflow." Reinhard agreed, but said "with the worst case view it's shure the balance doesn't look better than the real future outcome will be - it might probably look worse - but that's actually the point of a worst case view :)" .
12. Secure connections using ssh in GNUe
1 Nov 2002 Archive Link: "[IRC] 02 Nov 2002"
Topics: Common
People: Charles Rouzer, Derek Neighbors, Calum Morrell
Charles Rouzer (Mr_You) asked "how can you do secure db connections" with GNUe? Derek Neighbors (derek) said "we have full security wrapper now - i.e. plugin - that you can use kerberos, pam, etc" . "currently we use ssh" - "so the client is on my machine i ssh tunnel port 5342 to the internet machine postgres is on" . Charles asked how this would work with client PCs running Microsoft Windows. Derek said they could use "putty" which was a free ssh client written native for Win32 - cygwin was not needed. "we will likely add this as a plugin to the framework - so it will create tunnels everything on fly for you" . "the idea would be to have this wizard be in the framework and then execute the script on demand - instead of have to run externally" . Charles noted "you guys run GNUe locally.. but there are TONS of small business people that would rather run their small biz software remotely and not worry about a server or anything.. its harder to sell a "hacked" solution like ssh tunnel until its seamlessly integrated..." Calum Morrell (drochaid) asked "err, how is an ssh tunnel "hacked" in any sense? it's a fairly standard security consideration" . Charles said "my definition of "hacked" in this case means.. to the end user (newbie).. it may seem "hacked together" for them.. if they have to install more than one package and configure this and that.. rather than just configuring a GNUe Forms client." Derek said "the ssh stuff would be IN the client - i.e. to a lay person it woudl be click button enter username and passowrd - nothing more nothing less" . Even now, "it is install gnue install putty and click such and such to configure remote server done" - "if you were 'deploying' you could actually precreate the script - and they wouldnt do anything but connect" .
13. Documentation on the GNUe website
1 Nov 2002 - 2 Nov 2002 Archive Link: "[IRC] 02 Nov 2002"
People: Calum Morrell, Charles Rouzer, Derek Neighbors, Jason Cater, Peter Sullivan
Calum Morrell (drochaid) asked "what's the documentation side of GNUe doing recently?" Charles Rouzer (Mr_You) said "jason has a nice developer doc - its a doc for wannabe GNUe Application Developers ;-) like me ;-) - not python coders." Derek Neighbors (derek) said "its in jasons home directory i think - as its a work in progress"
The next day, Jason Cater (jcater) asked "have you given any thought on how to handle our new GNUe/Small Business Edition" for the website. "I think derek and I were leaning towards keeping it as GNUe (and so having it be a part of GNUe's website, somehow) - we just wanted a separate CVS tree" . Peter said "I think I would just add gnue-sb to the "Applications" section of the site" , noting that registering the project in CVS at savannah, the FSF's free equivalent to sourceforge, would automatically create a web page for the project on the www.gnu.org website. Jason said "we are **strongly** considering going to this gnue-??? model for all our gnue stuff" with seperate CVS repositories for "gnue-forms, gnue-reports, etc" . Peter said this "gives us an auto structure for the web site too - as each gnue-foo will have its own mini site on www.gnu.org - with /software/gnue/ as the official entry point" . Jason said "actually, that is one of our hangups - as we were expecting it to be harder to do the website :)" However, "this would let gnue.org be more of a community site like derek's always clamoring for - as we would always have easily accessible tool sites" Peter said "well, www.gnuenterprise.org already has things that we wouldn't ever put on www.gnu.org - e.g. the irc logs" .
Calum asked whether the website should "give documentation a much more prominent position?" Peter said "Documentation is a funny one - as we only have 1 docs page at the moment - but there are at least 3 levels - a) end user docs, b) application developer docs (people designing apps with gnue), c) GNUe developer docs" . Calum said "I always find it's one of the most important link .. and just now it's the 2nd last item on the menu, or into community and the 2nd last item, there" - he "like to see "GNUe Docs" beside "GNUe Community" in the top header" . Peter said that "Short term, I may as well go back to the old "fudge" - which is to have Docs listed in both Project (loghically just under "Downloads") - and also in Community as well" until there were enough documents of different types to warrant two pages. Calum suggested adding another major category down the left hand side for Documentation, and adding a tab across the top. Peter noted that the tabs across the top related to the three bullet points on the Home Page about "What is GNUe?" - Tools, Applications and Community, but he supposed he could "say that GNue was 4 things, one of which is a repository of docs" .
Jason was not keen on "another tab across the top - I think we have about as many tabs as we can realistically put there" . Instead, "I think each tab deserve's its own doc" page - "if we have a centralized doc page I think it's going to be so confusing" . Calum felt "if someone isn't sure which category a doc is likely to be in, could piss them off trying to find it" . Jason outlined how his idea would work - "the Forms/ page would still list all the form-related docs - but from the main menu, you could quickly get to all tool-related docs too" .
Jason said "one of the things we've been emphasizing over the last month or so is that the tools get lost in the apps - so we're trying to point out that GNue is 3 projects in one. I just think the documentation should follow in that focus as well - and I'm talking at a minimum here" . Calum said "whereas I would always classify documention as a separate project, as without really good docs, the tools/apps are useless to the less than technical users" . Peter felt "we need a lot more docs before any of this matters - but that will come" .
14. Bayonne and GNUe
2 Nov 2002 Archive Link: "[IRC] 03 Nov 2002"
Topics: Bayonne
People: David Sugar, Derek Neighbors
David Sugar (dyfet) said "we are close to a 1.1 release in bayonne" , the GNU telephony software project. Derek Neighbors (derek) needed to "get a frelling copy of bayonne working - i just got a new phone that does AOL" Instant Messaging (IM) "and im enamored - /me is drooling over possibilities" . David said "actually I have considered adding im support to Bayonne....so it can send sms messages and popups with incoming call info" - "well, it could though something like jabber, i suppose, where there are appropriate im/sms gateways. thats why i want to build an im module in bayonne" .
Derek said "hmm it does sms - grr i REALLY need to get a hold of it. the biggest hold up (i know you guys hate hearing this) is it needs to support normal modems - currently the hardware cost is way too high barrier to entry for 'free software' developers" . David said "free is not free as in price :)" . Derek said "i know freedom is expensive - just free software developers unless commissioned directly to install phone software can not afford 500+ for a professional phone card. i think if it ran on modems you would have a huge user base - i know for myself i would be running at home" . David said "I would like to get a modem version running - there even is a start of a voicemodem driver, but nobody has been working on it" .
Derek said "i have 3 clients that are considering buying hardware - but its hard for me to give strong recommendation when i havent driven it myself" . He had been working with some Bayonne people to get it set up for a client "so that i can get a bayonne install and get an idea for capabilities - and MOST importantly start writing applications to it :)" . David agreed - "and most important, I think, writing applications that integrate into/relate to gnue apps...." .
15. Property Inspector/scrollbars bug finally fixed
2 Nov 2002 Archive Link: "[IRC] 03 Nov 2002"
Topics: Designer
People: Jason Cater, Calum Morrell, Peter Sullivan, Calum Morrell
Jason Cater (jcater) noted that Monte Lin "who uses the debian packages has sent in a patch which fixes the Property Inspector issue :) :) :) - the one we've been trying to figure out for a year now - and it's a one-liner" . Calum Morrell (drochaid) asked "is that the "can't edit field with scroll bars present" one?" as mentioned in several previous threads. Peter Sullivan (psu) said "this is the great thing about free software - users who fix their own bugs - gotta love 'em" . Jason noted that he had previously claimed this "was a WX bug, and not a GNUe bug" . Peter suggested several possible lines of defence for Jason - either "the wx docs weren;t clear" , "no-one else noticed it either" or "derek told me to do it that way ;-)" .
16. GNUe and DotGNU - GNU-RPC and AppServer
2 Nov 2002 Archive Link: "[IRC] 03 Nov 2002"
Topics: Application Server, Common
People: Anthony Liguori, Andrew Mitchell, Jason Cater, Derek Neighbors, Jan Ischebeck
Anthony Liguori (aliguori) "had a quick question about the security adapter layer... noticed that ldap and active directory were connected to it (presumably, example security plugins) - ldap doesn't seem like nearly enough for enterprise security though (and security is a huge part of the app server) so i was wondering if this project is going to have a huge dependecy on active directory? or are there plans to work on an enterprise security layer?" Andrew Mitchell (ajmitch) said GNUe "probably plans to work on something - i very much doubt that there'd be an AD dependency :)" .
Anthony asked "is this the same app server that the dotGNU project refers too?" - "the one that supposed to support both j2ee and clr" architectures. Andrew said not - "the GNUe appserver is written in python" , which was "a programming language - we build the architecture :)" . Jason Cater (jcater) explained "our main goal with gnue appserver is centralizing business application rules - business logic, if you will" - "note that we aren't building a generic, all purpose application server" . Derek pointed to his notes from a previous meeting - "you will notice goal number one precludes use of j2ee and/or .NET" .
Anthony surmised "it seems that the GNUe app server is based on RPC, LDAP, and Python. Seems a bit like Active Directory..." Derek disagreed - if you had to compare GNUe AppServer to any Microsoft product, the closest was probably Microsoft Transaction Server, "and even there its a bad analogy" .
Derek said that, as of time of writing, GNUe and DotGNU were two very different projects, but there had been discussions on working together in two areas. The first was GNU-RPC - "also there was talk by dotGNU folks to look at appserver as something they are interested in" as a basis for their own web-based application server. Andrew, as "the main person who has feet in both projects" , said that "the gnu-rpc implementation will be getting worked on next week" - "and appserver will be used" by DotGNU " where it suits - if" GNUe's "appserver is not suitable for more generic tasks, then we'll sort something else out :)" . He explained "gnu-rpc is an abstraction layer to allow multiple transports to be used" (SOAP, CORBA and XML-RPC) - Derek explained "think perl-dbi only for rpc mechanisms - again its written in python" . Andrew said he was planning to write a version of GNU-RPC in C# for DotGNU. Derek said that "jcater is one that wrote gnu rpc" for GNUe, with the fine-tuning done by Jan Ischebeck (siesel).
17. wx splashscreen timeout bug
3 Nov 2002 Archive Link: "[IRC] 04 Nov 2002"
People: James Thompson, Dmitry Sorokin, Jason Cater, ra3vat
James Thompson (jamest) was "testing a fix for the wx splashscreen timeout bad window error" , as discussed in Issue #53, Section #8 (24 Oct 2002: Problems with Forms splashscreen) , "and think he figured it out - but I need to really load this machine down to test properly" . Later, Dmitry Sorokin (ra3vat) asked "does navigator work?" - he was getting an "AttributeError: GFUserInterface instance has no attribute 'splash'" . James said helpfully "I think the answer to your question is "yes, navigator is broken" :)" . He "had two options here - 1 ) keep the splash on top for 3 seconds - 2 ) drop the splashscren the moment the form was ready for processing - it's 1 line of code either way. I did 1 as on a fast system the splashscreen wouldn't have been up any time at all - but what would peopler prefer" ? Jason Cater (jcater) pointed out that "navigator is supposed to always suppress the splash" , but in Forms, he would prefer the "3 sec" option.
18. Stability of GNUe in production environments
3 Nov 2002 Archive Link: "[IRC] 04 Nov 2002"
People: Jeff Bailey, Jason Cater, Derek Neighbors, Andrew Mitchell
Jeff Bailey (jbailey) asked whether GNUe could be used "for a simple PO system. I think it's all good for that, yes? Like it's stable, generally won't eat data and all that? Backend would presummably one of our Oracle DBs or mysql or something like that. Clients are all win32." Jason Cater (jcater) said "haven't had it eat data yet :) - I think it'll be fine - as long as you don't have a user named" Derek Neighbors - "if that's the case, everything seems to fall apart" . Derek (revDeke) defended himself - "im not sure its EVER eaten my data - its not accepted it before" with some of the more unstable CVS versions "but never eaten it" . Jeff asked how the screens were accessed. Andrew Mitchell (ajmitch) said "screens are either supplied locally, or can be grabbed over http" . Derek clarified "they can be grabbed over any transport supported by python - http/https/ftp/etc" , using python's urlOpen() function or equivalent.
19. Checking user input before allowing a page switch
3 Nov 2002 Archive Link: "[IRC] 04 Nov 2002"
Topics: Forms
People: John Lenton, Jason Cater, James Thompson
John Lenton (Chipaca) asked "is there a way to abort a page switch? i.e. the guy clicked on a different page but he left this one in an illegal state" . Jason Cater (jcater) said "you could do a form.setFocus(some widget on the bad page)" . John reported "form. setFocus(widget) changes the focus back to the page, but it then changes back to the page the user clicked on - i.e. with debug-level set to 5 I get one "changing focus" when I clicked on the page tag, and two when I press "Cancel"" . He asked "shouldn't the pre-focusout trigger get called *before* the actual switch of focus?" James Thompson (jamest) agreed.
John asked how to cancel a page flip, "without writing one trigger per page" . This would allow the form to check user input amd "go back to the page the user was on if she forgot to fill something in" . Jason Cater (jcater) said "the goal was to allow triggers to return a true or false based on success/failure - so if a failure was returned everything stops" . John said "yeah, but the page has already changed, even" at the point the pre-focusout trigger was called. Jason said "then that's a bug w/page changing" . John expressed some frustration with the wxpython GUI toolkit, but recognised why GNUe used it, as it allowed cross-platform GUI so easily. "the worst bug is the widget-seems-to-get-focus-but-doesn't - that confuses the hell out of people" . This was "the one where you point at an entry but just on the border so the cursor isn't acutally I-shaped, click, and the | cursor appears in the entry, but the focus isn't there - it's amazingly simple to trigger that one by accident" . Jason said he had never seen this.
20. Next releases and their impact on AppServer
4 Nov 2002 Archive Link: "[IRC] 05 Nov 2002"
Topics: Application Server, Common
People: James Thompson, Reinhard Müller, Jan Ischebeck
James Thompson (jamest) said "for all the appserver developers in the house - jcater and myself are wanting to do a normal release of common, forms, reports, designer - then do what I've been calling a spitshine release at (0.5.0). meaning the next release is 0.4.1 - then 0.5.0 comes out with no new functionality - (I'm talking common version numbers btw) - we've instead focus on code cleanup, docs, bugfixes" - "0.4.1 would be basically cvs that we have now" . However, "how would this effect appserver? do we need to hold off on a common freeze for a while to get some features needed in there" ? Reinhard Müller (reinhard) felt "it would be better to do both 0.4.1 _and_ 0.5.0 of common _before_ " the Application Server team started "breakage" by adding the features they needed to Common. "However it would be good to ask" Jan Ischebeck (siesel) "too - because he's exactly the one that can answer this question - he always was the personified interface between common and appserver :) - and at our meeting he talked about some points he wants to change in common - however i didn't get the feeling that they were _urgent_" . James said "ok, the main reason I ask is we don't want the appserver needs to not get met - and I'd hate to try and say "no new features" if it hurts appserver. I'd rather do a series of 0.4.x releases to get what you require in there first - we've kind of promised ourselves no common|forms|designer 0.5.0 without good docs. Well, one thing that I hope to get into 0.4.1 is autodoc support for triggers in the base apps - so that it'll be easier to create them for 0.5.0 :) autodoc = our command line interface to dump various info about things like parameters" .
21. Debian packages into sid (unstable) distribution
4 Nov 2002 - 5 Nov 2002 Archive Link: "[IRC] 05 Nov 2002"
People: Derek Neighbors, Jeff Bailey, Jason Cater, Andrew Mitchell, James Thompson, Reinhard Müller
Jeff Bailey (jbailey) started to re-build the Debian packages for GNUe, so that he could get them into sid (the Debian unstable distribution. Derek Neighbors (derek) asked "can you commit them to cvs" ? Jeff suggested some possible changes to the CVS structure to make this easier - "The way I do the packaging is that the upstream source is just a tarball sitting in the deb. That way upgrades are mostly just replacing the tarball. I will cheerfully keep that in cvs if you'd like - that would be easy enough - but I'd make a new toplevel "packaging/debian/PACKAGENAME/" probably..." Jason wondered if his original Debian packages should have been marked as for woody (as of time of writing, the Debian stable distribution), "as they rely on python 2.1 stuff" . Jeff found and Jason fixed various bugs and other oddities in CVS (such as image files that had the executable flag set). Jeff explained "I'm taking a tarball of the CVS tree and stuffing it into my packaging. Which is mostly jcaters packaging." Jason said "I actually thought you weren't supposed to package against a cvs tree directly - that's why I worked from the source tarball" of the 0.4.0 releases. Jeff said "It depends on the maintainer. I only package against CVS when I'm friends with upstream." Andrew Mitchell (ajmitch) said it depended "and what level of masochism they like - i package from cvs for DotGNU stuff - because i am part of upstream :)" .
Jeff finished packaging GNUe Common, and asked "Why did you call the package gnue-forms-wxgtk instead of just gnue-forms?" Andrew said there were other Forms clients available, including php. Jason added "there's also a gnue-forms-curses and a gnue-forms-gtk2" which they would probably look to do Debian packages for once they were a bit more mature. "there probably should be a gnue-forms-base or such - as a good chunk of the code is the same" . James Thompson (jamest) agreed - "then driver .debs" for each seperate User Interface (UI).
Jeff could not find the man files for Forms in CVS. Jason explained "that's created dynamically when we package up a tarball" , but could also be generated on demand with the --generate-man-page flag. James noted that the man pages were not "100% complete" . Jeff said "Even if they're partially complete that's fine. The GNU project doesn't support man pages, right? =) So having a partial one, and a note to look at the texinfo documentation is usually fine." Jason said "as soon as I learn texinfo we'll hve a --generate-texinfo-page :)" flag as well.
Testing the packages, Jason noted "it runs - but my intro.gfd shows no labels - but running against my latest cvs copy, they do show." "I don't see how this could have *anything* to do with debs - but its odd" . Andrew Mitchell (ajmitch) said "don't try apt-get install gnue-forms gnue-designer - unless you want the 0.1.1 version :)" - "does gnue-forms-wxgtk replace gnue-forms?" Jeff said "replace, conflicts, yes." . Andrew said "gnue-appserver had best replace geas" . Jeff asked "Is there any value in doing appserver yet, or should it wait?" If the only reason was to replace the old GEAS (GNUe Application Server) package, he could "just file a bug and have it removed."
Later, he announced "gnue-common, gnue-forms, and gnue-designer were all ACCEPTED into Debian as of about 30 minutes ago. They'll be in sid tommorow. The geas update will probably wait until next week."
The next day, Jeff asked "Does appserver actually do anything yet?" Reinhard Müller (reinhard) said "it passes all data requests through to the database - and has a very very basic function for calling methods" . Jeff said "Okay, cool. I just want to make sure if I go throught he effort of packaging it that it's not a waste of time." Reinhard said "not sure about that - it does nothing "useful" now - it's more like an empty box that can be filled afterwards - the only reason i see it would make sense to package it is for "marketing" reasons" . Jeff agreed - "I'm not so concerned about the usefulness. I just want it to somewhat work as advertised."
22. Web-based solutions for GNUe
4 Nov 2002 Archive Link: "[IRC] 05 Nov 2002"
People: Jan Ischebeck, Jason Cater, Jeff Bailey, James Thompson, Derek Neighbors
It was asked if there was a php Forms client. Jan Ischebeck (siesel) said there was, "but not very stable at the moment." . Jason Cater (jcater) sighed "I guess I'll never understand this web-based frenzy" . Jeff Bailey (jbailey) said "Zero footprint clients save us a pile of cash here." Jason said that was different - "I love the thrill of 0 footprint too" - "I squeal like a stuck hog" . Jeff mused "I wonder sometimes if I'm happiest having never met any of you. =)" James Thompson (jamest) said "no need to wonder, you are" .
Derek Neighbors (derek) said "there is no such thing as zero footprint client - web apps need a broswer ive yet to find one that is ZERO footprint - to top that off they are all different - and EVERY corporate one we have looked into MANDATED" Microsoft Internet Explorer 5.0 or higher. He was "thinking at some point of making a 'wrapper' for forms so its a plugin - so you can run forms like 'applets' i.e. you go to a url that uses forms - and it tells you to get gnue plugin v0.4.1 - you go and grab that and voila now forms works. The sad thing is it will quell the 'web freaks' - but in essence it is nothing more than normal forms wrapped" .
23. Contributors to GNUe
4 Nov 2002 Archive Link: "[IRC] 05 Nov 2002"
People: Jason Cater, Andrew Mitchell, Daniel Baumann, James Thompson, Jan Ischebeck, Bajusz Tamás
Jason Cater (jcater) explained that "my problem with roadmaps is - it doesn't define where we're going - it defines what others expect me to do" . Andrew Mitchell (ajmitch) said "there are others who can help!" but Daniel Baumann (chillywilly) said "nah, no one is worthy of hacking jcater's code" . Jason said "I welcome help - and I'd like to take this opportunity to thank those who have started coding more and more" , including Bajusz Tamás (btami), Andrew, Jan Ischebeck (siesel), the members of project Papo - "I *know* I missed someone... so don't take it personally" . James Thompson (jamest) sniffed "now we know how he feels about my code" . Jason said "you haven't started coding more and more - you've always coded" .
24. GNUe Project organisation - version numbers, roadmaps, testing and branches
4 Nov 2002 Archive Link: "[IRC] 05 Nov 2002"
People: Jeff Bailey, Andrew Mitchell, Jason Cater, Derek Neighbors, James Thompson, Daniel Baumann, Matt Rice, Jan Ischebeck
Jeff Bailey (jbailey) said "The version number in Debian is going to be 20021105." Andrew Mitchell (ajmitch) was not keen on this, "unless this is the 2million, 21 thousand, and fifth debian revision :)" . Jason explained that, for help desk purposes, the About box in the CVS version of Forms said it was 0.4.1a ("a" for "alpha") - "but it will become 0.4.1 when released" . It was noted that Debian - and other packaging systems - would regard 0.4.1a as a later version than 0.4.1. Jason said that it was not really possible to change this before 0.4.1 was released - James Thompson (jamest) suggested going straight to 0.4.2.
Derek Neighbors (derek) said the project "might as well release 0.5.0 - we have no method right now (seriously) - other than chaos - its completely abitrary" . "there is no 'roadmap' that states what release is what - so there is no control - at which point who cares if its 0.4.1 or 0.5.0 (independent of the alpha question) - there is nothing stating what features will be in what releases - so USERS have no idea until something is released what is what" . Jason said "we have our internal markers - a lot is dependent on whether common's API changed - if common's API is vastly different, it has to be a major point release" . Derek said "until now its been no big deal - as we didnt have mass of other developers - and we didnt have production users" . Jason said "it's not perfect, nor well planned - but it's not arbitrary" .
Derek said "we need to very soon get to the point where we really maintain 2 releases - so that someone on an old release can expect to get their bugs fixed - but not expect a bunch of expiremental features or major breakage" . James said "i thought someone was going to backport patches if we tagged cvs?" Derek said "that would be me - my problem is cvs commits are going like mad - but the cvs commit messages frankly suck - and there is no roadmap or system used to manage what should be backported. My proposal is we have a strict feature roadmap for the new releases and only if some feature is minor in nature does it get backported" .
James said "i can't see that working with free labor" . Derek emphasised "i never said 'put dates on features'" . James explained "what I'm saying is that like tonight - i finally got sick of the no query by detail - so I'm fixing it. This isn't on any roadmap - other than I'm sick of it" . Derek said he was "not saying that roadmaps cant change" - the importance of the roadmap was that it provided a place to park feature requests for the future rather than the next immediate release - "at some point one has to say this road is closed" - creeping featurism had been had been a factor in the length of time taken over the last two GNUe releases. This had affected GNUe's relationship with other projects, including papo.
Also, with a roadmap, it would be possible to split changes in CVS head between bug fixes to existing features (which would get ported back to interim 'stable' releases) and new features which belonged only in the new release. James said "nothing prevents that today however the API is constatnly in flux, we are a 0.5.0 project - I can see something stable as we approach 1.0" . Derek said "as we get more developers and more solid releases it will be mandatory to avoid chaos" .
James said "well, another way to approach this problem that i would prefer - an (semi-) automated testing framework" . Jason said "I would too - and I've BEGGED people to do it" . Derek agreed "certainly that is necessary to help speed moving a release out the door - as thats one thing that has made the process difficult - feature freeze is called - and it takes too long to test to point where its too tempting to add new features in" . Jason noted that Jan Ischebeck (siesel) "has even taken the initiative to start one in samples/testcases - but I don't see testcases flowing in" . James said he had "been adding to it for what I'm working on - but it's not standardized yet - was hoping to make it part of the 0.5.0 cleanup release" . Daniel Baumann (chillywilly) noted "there's different types of testing - black-box, white box, etc. - white box requires good knowledge of the code that is being tested - the *internals* of it - there's probably only 3 ppl right now that fit this description, imho - with perhaps a couple that have a decent understanding" . James explained "i wanted to make a gnue navigator thing - with various forms that had pages lableed step 1 2 3 4 - and a final pass/fail step" . Daniel said "basically we need a testing team and/or go intoa testing mode" . James felt that "anything is better than what we have now - which IIRC was jcater saying "prereleases is here please test" - then have people test after the final release" .
Daniel said "soem ppl can't code without some structure as they like to see down the "road" ;)" Derek said "the people closest to gnue have the roadmap in their head - and cant see why anyone is frustrated that there is no structure - because when you are close to it, there appears to be structure (and there is) - just its not in a format that can be well shared" . He was "not so much requesting things be radically different than they are now, just that they be better presented to the public" . Jason, referring back to Issue #52, Section #7 (19 Oct 2002: Frustrations with Release Process) , said "we are already stretched pretty thin - the few of us" . He felt it was unfair to suggest that the core developers were not responsive to end-user needs - "I've spent a HELL of a lot of time on the developer's guide" , which he personally had no need for. Derek said "im willing to write up roadmaps and maintain a back branch - but not if i have to swim against the river to do it - i.e. if you and jamest loathe the idea of roadmaps i woudl rather not try to document whats in your heads for the public :)" . Jason alsoo objected "you make it sound like we just only add new features and never look back - jamest and I take a release every so often - and do NOTHING but cleanup and profiling" , as planned for the 0.5.0 release. Derek said "actually my complaint is you guys do far too much clean up :) - and my hope is to get to a point where others can 'clean up some' and you guys can add new cool stuff" . Jason agreed, but said "I would also like some donuts - but I don't have any - so it won't happen" without extra volunteers. He felt over-managed on the project. Daniel said "I think all FS projects could use a bit of management, however, I don't think they need "managers" ;)" . Matt Rice (ratmice) suggested "I think we need the man with the yellow hat from curious george books" .
25. Primary keys in GNUe Schema Definition (.gsd) files
5 Nov 2002 Archive Link: "[IRC] 06 Nov 2002"
Topics: Common
People: Ariel Calò, Jason Cater
Ariel Calò (ariel_) said "i've seen the new documentation of schema (gsd) - we have a type 'key' (old a uto) that is translated to a db specific type, may be serial, int4 or whatsoever. Now a primary key has 2 functions: 1) enforce uniqueness of rows - 2) be referenced by foreign keys of other databases. Suppose i have a table with a field that references the primary key of another table - i want that field to have the same type of the pk in the other table. If i declare it of type key may be this is translated to serial (that i don't want since it is only a foreign key). i cannot declare it int4 (or similar) since i don't know how key is translated. i think a solution is another type fkey that is automatically translated by the specific driver to whatever is appropriate for the specific db backend" . Jason Cater (jcater) pointed out that "key != serial - there should be an auto="" that makes it serial. So in the primary table you might have mykey --> type="key" auto="y" - then in the referencing table pointerToKey --> type="key"" .
26. Bugs in Microsoft Windows
5 Nov 2002 Archive Link: "[IRC] 06 Nov 2002"
Topics: Forms
People: Bajusz Tamás, James Thompson, Jason Cater
Bajusz Tamás (btami) reported a bug with GNUe Forms on Microsoft Windows 32 - "with tabbed forms - when changing pages with mouse - after that going to the next record - the "content" doesn't refresh" . He pointed to the area of the code which would mean that "changing tabs with mouse will be OK - but PgDn/PgUp wrong - but it's only on win32 IIRC" . James Thompson (jamest) suggested a possible work-around, but said he "has no working win32 boxes handy" to test this himself. Bajusz said "i feel nobody likes to fix win32 bugs here :)" James said "i started hammering thru the dcl bugs last nite - so if they are in there I'll make as attempt to fix them" . Jason Cater (jcater) explained "it's not so much we like to fix non-Win32 bugs vs Win32 - it's that we don;t have win32 machines handy :(" Bajusz said "i think fixing win32 bugs is important, cose most of GNUe users will use win clients - while M$ is so strong in offices" .
27. Adding a CSV 'database' driver for Integrator
5 Nov 2002 Archive Link: "[IRC] 06 Nov 2002"
Topics: Common, Integrator
People: Jan Ischebeck, Jason Cater, Andrew Mitchell, Jeff Bailey, Derek Neighbors
Jan Ischebeck (siesel) asked "does anyone of you have an urgent need for integrator?" Jason Cater (jcater) said "relatively so" - he had "already started on it - I think my latest stuff is in cvs" . He added "one of the biggest things I'll need to do next (That I haven't even started on) is a CSV / Flatfile dbdriver - even a simplistic one - as my plans for integrator are primarily mapping from source datasources to destination datasources" . He explained "this isn't actually to run apps against - I need to be able to load CVS files into my database and dump them from my database" . He "certainly wouldn't plan on running forms against a CSV file, although it would be possible I suppose.... :)" Andrew Mitchell (ajmitch) said "i can try & start on one tomorrow if you'd like, not sure how you'd declare field names & stuff like that tho - as it would be useful for me as well" . Jason said "that's part of my hangup - I'm wondering about having a child node of datasource" , where the flat file defintion would be included as part of the datasource definition - "but am not sure the best approach" . Andrew suggested "for something quick & dirty, just name them 'column1' or 'field2', etc - would allow for conversion" . Jan said "if the flatfile dbdriver is just used for loading csv into integrator, why should we define special names for the fields - wouldn't a "field_1", "field_2" etc. be enought?" Jason said "I don't think it will necessarily be 1 way - but that's fine too - field_1 would work both ways" . Andrew said "some CSV files may have headers iirc" . Jason said "most loading software I see have a "skipFirstRow" option" . Andrew said "yep - that first row may contain field names if you're lucky" . Jan "would like to put all that stuff (skipFirstRow, delimiter....) into the connection.conf file" . Both Jason and Andrew were keen on this. Jan said "what about writing? Should read+write access be supported. (that would mean to rewrite the file after a COMMIT)" . Jason said "I can see two options 1. skipFirstRow=0|1 or 2. useHeadersAsFields or soemthing like that - (they don't have to be exclusive)" - "actually, skiprows shouldn't be boolean - it should be an integer - as I've gotten some funky stuff frm vendors before - esp. when they aren't using a real application but are instead doing in a spreadsheet and saving as csv" . Jan suggested "what about "firstRow=skip|fieldnames|data"" - Jason added "and default to "data" ?"
Jan noted "IMHO the Integrator Proposal speaks of converting data from a datasource into XML, move it into a RULE ENGINE and export the XML into a datasource again. the RULE ENGINE is quite similar to a XSLT tranformation for me. So I would like to add a module, which does kind of XSLT transformations, but is written in python, to common." Jeff Bailey (jbailey) asked "What's wrong with regular XSLT? Use an XSLT namespace.... Like, if it's presenting as XML, you could just use <xsl:*> tags to embed commands in." Jan said "I would just like to fake XML to the RULE ENGINE, because I don't think that it is good to convert huge database tables to XML just for fun." Jason said "reports is a rule engine" . Jan asked "so you would recommend to take code from reports to write integrator?" Jeff suggested, rather than re-use code from reports, Jan should "move code into common and use in both reports and" integrator.
Later, Derek Neighbors (derek) said "i personally wouldnt make it as complex as is talked about (well to a degree) - as i think there is a difference between 'integrator' and 'csv driver'" .
28. DCL as a bug-tracking system
5 Nov 2002 Archive Link: "[IRC] 06 Nov 2002"
Topics: DCL
People: Jeff Bailey, Derek Neighbors
Jeff Bailey (jbailey) noted that the gcc (GNU C Compiler) project's mailing list "seem to indicate that folks there want to switch from Gnats to Bugzilla" as their bug-tracking system. "People (and by this, I generally mean Derek) who would prefer that they use DCL should probably speak up." Derek Neighbors (derek) admitted "im a dcl evangelizer - BUT its more 'work order' oriented than software bug oriented - i.e. there are several features missing that would make it better" specifically for bug tracking - "but i think it does pretty darn good" . He felt that "with most software you hope you are doing more new work than fixing bugs - and i think DCL is FAR superior in task management - but weaker in bug management (though it does an ok job of this). If 'versioning' stuff was beefed up a bit it would be pretty good - bugzilla is great for version management, but its really bad for end users to deal with (imho) i.e. its daunting to deal with unless you use it religiously" . Jeff said "I know several people with the same complaint."
Sharon And Joy
Kernel Traffic is grateful to be developed on a computer donated by Professor Greg Benson and Professor Allan Cruse in the Department of Computer Science at the University of San Francisco. This is the same department that invented FlashMob Computing. Kernel Traffic is hosted by the generous folks at kernel.org. All pages on this site are copyright their original authors, and distributed under the terms of the GNU General Public License version 2.0. |