Scott Lamb reported Logging into to PostgreSQL from GNUe Designer 0.1.1 on win32 does not work well for me. It was using his Windows login to try to log into the database server, instead of the database user name he had given it. Derek Neighbors reported It is fixed in CVS. You can grab a snapshot from http://www.gnuenterprise.org/downloads/. I believe we plan on releasing a 0.1.2 soon because of this very bug.

Later, on IRC, Derek Neighbors (derek) asked slamb: you get things working? Scott (slamb) confirmed yeah, it logs in fine now with the 29 Jan 2002 snapshot :). James Thompson (jamest) confessed we can't blame the pypgsql driver as derek stated - it was my bug in our pypgsql interface . He added that there was still a bug with editing properties - if you have a scrollbar on the property editor then you can't edit anything He said i was under the impression this was a wxwin bug - but I noticed on Fri that the wxpython demo doesn't suffer from this - i hope to look at it in the very near future.

Scott asked how are those win32 builds made, anyway?. James said you do a complete install of all dependencies on windows - add pypgsql, mysql, and pywin - then run it thru mcmillian installer version 4.x - then use inno to package it up.

Later, Scott reported made new form, querying doesn't show anything. It adds new stuff fine, though. It turned out that the keyboard shortcuts for enter/execute query (f8/f9) were still working but the menu entries just don't do anything.. James and Calum Morrell (drochaid) said they hadn't noticed this before because they always used the keyboard shortcuts. James said i imagine that the event system changed and the menu/toolbar entries didn't.

Further to Issue 5, Section 2, Daniel Baumann (chillywilly) reported that my ODMG book came in the mail today - "The Object Data Standard: OMDG 3.0". Andrew Mitchell (ajmitch) asked what are your GEAS plans now? Daniel said he wasn't sure - I have some material to digest first - but this ODMG book should help things a bit

Several days later, Daniel Baumann confirmed he was still working through the ODMG book, making notes on all the goodies in my book, which I plan to make available to some interested GNUe hackers....ummm within limits of copyright ;) Several people tried to guess what ODMG stood for, with varying degrees of obscenity. Daniel explained he was reading about metadata and the ODL Schema repository - in hopes of making a GEAS that doesn't suck, but felt the ODMG metadata stuff is confusing. Jason said what d'ya expect with objects? something simple????. Daniel rretorted plz - objects rock - it's not that complicated - I dun figured it out. The intention was to avoid having to design an object framework for GEAS from scratch, but just write a Free implementation of an existing standard. He thought that the fun part will be converting all those gcds and doing a proper design there ;P. Jason commented that I don't want to be around when you tell reinhard his gcd lexer is shit. Daniel said it's not - the lexer will provide a good base - what it parses things into might change though - we need some real objects - not hacked up andrewm C structs ;P.

He said that using ODMG would allow you to define schema to store the objects and then you put that schema in the tables of the relational db - you want to add a new object you just add to the "repository". This meant you could write it once and forget about the damn db and just use objects. This would take longer to design, but not if you just use an already made standard ;). He noted ODMG is pretty language neutral, you even do python bindings if you wanted to

Following on from and , James Thompson (jamest_) said i need custom menus soon (before the 15th of this month :( ). Daniel Baumann (chillywilly) asked has anyone looked at the XForms standard?. James said long ago - it didn't (at that time) fit forms target - and IIRC was pretty complex. Derek Neighbors (dneighbo) said i need menus too :) . He said navigator is neat and all - but it wont work for production stuff imho. He added i think the basis is right - its a matter of reading that file into a gfd as a menu - not a separate thing that hangs off the side. What he would like is gpd file loaded into existing menu at start-up. You could then decide whether to a. only have one form in the framework open at a time - b. allow multiple to be open - c. allow multiples but contain them like an MDI app all within the 'container'. He felt a or c make most sense - and a is probably easies to do right away and would perform more like traditional app andd make easiest for 'curses' or other limited UI's.

Later, Jason Cater (jcater) said he disagreed - having multiple forms open was equally valid. He said I'd prefer to see 'Submenu items' that are attached to 'triggers' and leave navigator as a separate tool, the way that Oracle and SAP did it. James explained what I was wanting to do in menu system - GMenu's make a tree - they have things like Text, HotKey, Trigger - so that all menu items fire triggers and that these triggers have default names so that a form could overwrite the std trigger with a custom one. For menu placement, menu tags in the XML needed to place themselves in the normal drop-down menus - find File.Save and put me after it. Because GMenu would be based upon GObj then the entire default system menu and form customizations would be exposed to the trigger system. He added my initial thoughts are to put this in common - so that eventually any gnue app could define it's menus this way.

Jason asked did you see the pic of SAP Easy Access screenshot someone posted? that's what I'd like to have ;). James noted I'm pretty sure that I don't want a toolbar entry that isn't on the menu so I think the toolbar system should just contain links to menu entries Menu items could have a toolbarPosition and toolbarIcon attribute that if filled add to the toolbar. This would mean that all toolbar entries must be on the menu, they can't exist without a menu pick. Jason agreed - imho a mouse is 'always' optional equipment :). James asked should we include the toolbar def as part of the menu def - or have a completely seperate GToolBar that sits seperate but acts to only "fire" the menu picks. Jason preferred the option that would also allow us to provide "custom" toolbars if that was desirable. James gave some examples to clarify what he meant.

Derek (derek) returned, and noted the sap thing is ok, if the current navigator was cleaned up to that kind of polish it might be pretty cool. He felt we will need ability to load gpd's as menus directly into the 'framework' - but i see value in a 'pretty' navigator as well. In his view, navigator is hard to navigate for complex menus :) - i.e. unless you know exactly where things are under the current navigator they are hard to find, although the current navigator is perfect for curses world almost untouched :).

Referring all the way back to , Jens Müller noted the reference to storing infomation from previous employers in the Human Resources module, and feared this might breech European privacy laws. Derek Neighbors thought this was a good point, but added I was not saying GNUe would or wouldn't have this functionality, I was merely asking for the specification submitted to be more clearly defined in this area. I am not of a Human Resources background, but I would be suprised if companies kept any information on prior job history on an employee other than hand jotted down notes from reference checking or letters from past employers , even in the United States.

Daniel Baumann reported on Some links that you guys might find interesting at http://www.omg.org/technology/cwm/ and http://www.cwmforum.org/. He felt they had Some pretty interesting stuff on EDI/Data Warehousing/Metadata exchange. This makes me thing of GNUe Intergrator, OLAP, etc.

Possibly rembering the discussions from Daniel Baumann (chillywilly) noted nickr is supposed to be writing something for Document Management for GNUe. Nick Rusnov (nickr) said maybe I should actially produce some docustore code - When I finish ratmaze and poplar I'll work on docustore. He added ratmaze is sort of like docustore actually - more like 'projectstore' He went on to say i just realized hov I can save a lot of time with diplicated work - i'll rewrite the docustore spec to use ratmaze's metadata structure - and then use ratmaze as a prototype. He explained I keep all my projects and data and such in a directory tree like proj/blah/blah - ratmaze basically adds metadata to this structure as separately stored xml files - also enables journal entries and thumbnails and such . This meant he could have 'homepages' for my projects and have them all browsable - maintaining the same directory structure as the projects and allowing other info to be associated with them.

Michael Dean (mdean) wondered if it could be implemented within DCL to some extent? Nick said his metadata is in xml files rather than database tables. Michael said dcl can attach files to projects, but it doesn't allow you to place a structure for storage into the project. Nick and Michael agreed to work together on XML or database table specifications. Nick explained What ratmaze will do is chase down all the constituant parts of an index and then transform them with an xslt using a cgi script written in python, and using DOM to manipulate the page as I assemble it - then using 4Front to transform it. He had chosen 4front over Sablotron simply becuase it had a Debian package.

Nick explained he wasn't working to any particular Document Storage model - I'm creating a system based on my experiences mostly, not on any document management system that currently exists. I should probably look at those. With this new idea to unify docustore and ratmaze I have a little to think about. He had an oldish proposal, pre-ratmaze and also some very old DTDs for the XML data, but If I want to merge all of this into one metadata management system, then I'll have to come up with some new stuff.

Later, he explained the docustore platform is more of a database engine for document management than a direct document management system.. something like phpgroupware or dcl or geas or whatever would call upon it for document storage needs. They would use GNU-RPC to communicate with the document store, which would enable them to use any supported RPC mechanism, not just CORBA. He said he hadn't progressed very far with his docustore ideas yet, but I'm good at doing things that theres cohesive demand for. The thread degenerated into a competition to see who could be the most demanding.

The next day, Stuart Bain (stbain) wondered now that I know python - how can I help?. He would send back his project assignment docs. Derek Neighbors (derek) said the architecture is highly usable in 2 tier mode - n tier doesnt play nicely yet - and applications arent 'shrinkwrapped' but several people use for production stuff. He said you could use GEAS today if you wanted to write your own front ends to it. He added i just want to make sure im not saying geas doesnt do anything - as it does. Daniel Baumann (chillywilly) pets poor geas - I still love you - don't listen to them - geas juts needs more lovin' .

James Thompson (jamest) said Stuart should pick whatever area he was most interested in - that's a weak answer I know. Stuart said it would be nice to have a system I can look at and say, "You know what? I can replace that monster AIX box w/ one little Linux PC that will run your entire operation." - or three boxes ;). He was probably more interested in applications than tools, but I want to know the ins and outs of the tools too - tools == means to end(apps). Nick mentioned this document management system that I was working on. James said stbain: if you could help nickr that'd be cool.

Stuart asked whether Nick had considered an imaging interface go go w/ that? That way you can have a document imaging and management architecture under one roof. Nick said he had envisioned having any 'object' able to be managed under the same storegae system. OCR would just be another input method.. Stuart said many Document Management systems he had seen were for HR departments - they are simply document repositories for the immense amount of paperwork companies have to process to keep up w/ their hundreds of employees - if they want to look up my docs, they punch in my SSN and scroll through the descriptions - they would love to have it integrated right into their ERP software but this was rarely possible as at time of writing. Nick said Its pretty trivial, really. the key is making it abstract. James noted you might be able to use a fair amout of common for database and RPC abstraction. Nick said this has been recommended to me - particlularly database abstraction stuff. James added we've also got a base app in there - that gives you debug levels, profiling, config file reader, the start of a trigger system - and lots of other little things .

It was asked whether there were any applications currently using the GNUe Tools. Jason Cater (jcater) said yes, but inhouse stuff, as discussed previously in . Derek Neighbors explained they were not availalbe under gpl :) i.e. custom solutions. Jason explained gpl covers redistribution and inhouse solutions typically aren't redistributed. There would be no advantedge to using the GNU Public License (GPL) if its an app written around a single business' idiosyncracies. He noted unfortunately, aside from our samples/ directory, all the real-world uses of forms right now are in-house stuff . However, early adopters are a GOOD thing :).

Calum Morrell (drochaid) said that some companies might be afraid that releasing customised GNUe applications under GPL might allow your competitors to see how you do it and potentially modify there own system with some improvements gained from knowledge of yours. Jason said I imagine its not as secret as most companies would like to think ;) Derek said i see most people will take the 'base' then extend and not give back - which is fine, though over time i think they will realize the computer system is only a small advantage.

Jens Müller (ICJ) said what is really important are different tax laws in diff. countries - they have an heavy impact on accounting. Derek agreed, saying i see us 'templating' things by region/ location to handle tax specifics and such, as previously discussed in and . He wanted to avoid having a system with a large number of tables, most of which were irrelevent for most users. But rather you get a small base then 'apply' things based on 'country, tax laws, etc etc etc). Jens said that would be fine, as long as there was good package management. Derek said you might have apt-get gnue-accounting, apt-get gnue-accounting-austrian-tax, apt-get gnue-accounting-euro-conversions or soemthing - only wont be apt-get - probably gnue-get or something similar.

Stuart Lamb reported a bug - When a commit fails (a referential integrity violation), it pops up a dialog box as expected. But apparently doesn't tell the database server to rollback - because anything in the future (whether I hit "commit " or "rollback" next, says this - libpq.Warning: NOTICE: current transaction is aborted, queries ignored until end of transaction block - and that transaction block never ends. This was with pypgsql. James Thompson (jamest) tested this, and confirmed that it could cope and recover from an attempt to duplicate a primary key, but not from a referential integrity error (i.e. a missing foreign key), despite the fact that they call the exact same error routines on our end.

Jens Müller (ICJ) asked Why isn't there a mailing list with CVS commit messages?. James Thompson said there was, at http://mail.gnu.org/mailman/listinfo/commit-gnue, but looks like it's not listed on our web site :( Jason Cater (jcater) thought that may be a good thing :). Derek Neighbors (dneighbo) explained fwiw: generally we hope only those really interested find it :) as admining a bunch of mail to commit is a pain :).

Jan Ischebeck (jan) confirmed he had continued the php client referred to in . It could now do a simple query on an database, input data into cache, marking as deleted, adding new rows, an About box, and ... some bugs ;-) James Thompson (jamest) said he had made some changes to the main Forms client this week, and we need to coord better so we don't break stuff on you.

Later, Jan confimed he had put a fairly limited sample on the web - not all functionality was working yet. Jason Cater (jcater) noted that it didn't work at all with the Konq web browser, but did with mozilla. Jan said he had tested it with Mozilla, but I m no MASTER OF THE JAVASCRIPT so it might not work with all browsers. He asked if anybody knows how to submit a form with an image button, ... and tell the server that it is submit button no. 1 and not submit button no. 2, in a way IE, konq, netscape 4.x and mozilla understand,... please tell me.. Michael Dean (mdean) offered to help. Jan was trying to catch an key event like "page down", but it dont work. - Just "ALT-P" and "ALT-N" are working now. Michael noted that key events aren't available for all keys :-( . He later confirmed as of IE 5.0, you can catch pg dn/up for onkeydown and onkeyup, but not onkeypress - if you're just wanting to catch when the press a key, you should use onkeyup - onkeydown will fire constantly while they hold the key down .

Scott Lamb (slamb) thought It may not be necessary to use javascript to do your graphical submit buttons. I'm looking at W3C spec. You can have multiple submit buttons and you can have graphical ones. It says "If a form contains more than one submit button, only the activated submit button is successful." Which I think means only its values are submitted, so you can know which was clicked. Give me a second and I'll actually try it. Michael agreed, but said if you need to change form data or the action based on which button, you probably need Javascript.

Jan felt he could now write javascript that would work on most browsers, Just lynx and links won't work... . Scott didn't feel this was too serious, but asked actually, what is it about lynx that makes it not work? I'm curious now.. Jan confirmed that, Although not working, lynx is not too bad . It degraded to [tb_save.png]-Submit [tb_new.png]-Submit ... so at least you had some idea of what was happening. Scott suggested putting in alt="" attributes for the images to improve this, which would also give tooltip hints on graphical browsers. However, Jan expected that lynx doesnt like to send x,y arguments for an image he cannot display...... so if you can click on the image in lynx it wouldn't work. Scott confirmed by testing that lynx is giving me .x and .y values of 0 on image submit.

He noted that lynx didn't like the encoding type in the web page header - the W3 standard, which it complied with, required something different. Jan said Thanks, i will never trust any SELF written HTML course anymore.. Scott agreed - yeah, most of the info on the web about html is pretty misleading. why I just go to the official W3 specs? Jan said he had loooked at them for something some time ago, and it was a little bit confusing, so I didnt use it. But looking on the HTML 4.0 spec. ... GREAT. I ve never heard about LABELS in forms etc.. Scott agreed, but noted small catch, though: several of the things on the W3 specs aren't really done by any browser.. Michael reckoned html form labels rock - don't make checkboxes or radio buttons without 'em ;-).

Jan asked for ideas about the rest of the UI. He explained In the beginning I had planned to make just the actual row in the actual block editable and now everything visible is editable. What do you think is better?. Scott thought that minimizing the number of times you have to submit/wait for the server to respond would be good - so from that perspective, the way you have it now is much better. Jan agreed, but asked about cursors - Without cursor, how do I know which row to delete (if i don't use javascript)? Scott thought how you have it now is probably the best you can do. No way you can know that w/o JS, so it'll always be a bit confusing to newbies. I like the way you highlight the deleted row in red. Actually, you could have a seperate delete button for each row. but that would require a change to the form definition.

Jason Cater (asked) can you think of some polls for our homepage? the current one is a little outdated He suggested some object vs relational vs flatfile database usage poll? or, what is your preferred DB backend? Derek Neighbors (dneighbo) liked the sound of this last one, and gave a list of possible answers. Jason said I would qualify that with "for your business processes" or "business data", etc - as a backend for a dynamic website is irrelevent to us, imho. Derek wondered if the question should be which database do you PLAN to use with gnue. Jason also wondered if we should do a "business size" poll to get a feel for the class of enterprises mostly viewing our web page. Calum Morrell (drochaid) gave some immediate answers for all the proposed questions, and asked do I win a prize?. Jason said the prize was 4 days/3 nights in the beautiful #gnuenterprise - (does not include travel or food) . However, Derek said it DOES include - free insults, complimentary fish slappings of you CHOICE, poor and crude jokes.