[Accessibility] Corrected minutes for 11 Jan 2006

John Goldthwaite jgoldthwaite at yahoo.com
Thu Jan 19 05:36:48 PST 2006

I've added Olaf's corrections:

 01/11/06 Accessibility meeting             

Gunnar Schmidt
Olaf Schmidt
Bill Haneman
Janina Sajka
John Goldthwaite
Cathy Laws
George Kraft
Randy Horwitz
Peter Korn
Peter distributed minutes from last week,. Olaf had a question relating to the adoption of KDE in Spain. Janina will incorporate a correction and post the corrected version.  

Janina transcribed the audio from the conference taking 52 man-hours.  This resulted in 15 hours of audio and transcripts that have been published on the website.  I am proposing to bill $5200 from the NSF funds for the transcription work. Does this seem excessive?
Earl- What are normal charges? It would be good to get an idea of what fair and reasonable cost would be.
Peter - It should be more than a normal audio engineer because of your domain expertise.  It is hard to say if it is fair rate or not, $100 doesn’t seem excessive, $50 seems too low.  Perhaps do a survey of what a couple of people charge and go up a bit because of your domain expertise.
Janina- Should I bring it back to the group?
Peter- do it in email.
Janina- will do that and post in an email. 

Janina- Earl, do we have any comments?
Earl- we have one request for comment logged from Olaf.  No outside comments. Peter took care of his action item on forwarding to the alias. I had communication with Norm Robinson.  Janina had an a W3C action item.
Janina- I did that. George did troll around looking for our announcement and it doesn’t appear to have been picked up anywhere.
Earl- I’ll send an email to identify and outstanding items, not much else to report.
Janina-Concern s  me that we can do some thing fundamental and not get news coverage.    
Earl- we talked about involving the FSG but that would be when the spec comes out.  I don’t think they can do anything now.

Bill- the main thing is that we get germane comments from experts during the comment period. I hope that we will get that from the list we published to. Are there other important lists that we haven’t reached?  Maybe we need to get a list member as an advocate and get them to point to our announcement.  
Earl- we’ve gotten on the one’s we identified.  
Janina- have we heard from Norm Robinson that he was able to get it out ?
Earl- he got it on his Section 508 alias list but he said it wasn’t all the 508 coordinators.  There doesn’t seem to be a official list.  Norm provided feedback on the spec earlier.  To answer Bill’s question, we did a good brain storming earlier and hopefully got all the important ones.
Bill- I’d be more worried if we get no publicity when we get the final spec.
Earl- that’s all for Lake Access X.

Bill- as an aside, for a while we were talking about the ABI library spec part of the keyboard spec.  A lot of that was in my lap and we pushed it off to phase 2.  How do people feel about that? I feel an ABI component would be important.  We had the feeling that spec’ing the whole of XKB was impractical.  Does anyone feel we need to accelerate the ABI component?  
George- I feel strongly in favor of having an ABI component.  To me that is important for the desktop to link to and telling the desktop what is stable in that library.  What can be used and what is in the header files and if anybody omits something or tweeks something, you know how to point the finger at.
 Earl- Bill’s got an important part of the keyboard spec but he has just as important part on the AT-SPI. We took a lot of Bill’s time on the keyboard access spec. My reason for not going for the ABI part of the keyboard spec is that it would be fairer for the rest of the effort if Bill could focus on the AT-SPI work.
George- you have everything you need for the specification, the ABI can be gleaned out, the header file information can be taken out, you have all the man pages. You have all the behavioral specification there. Everything is there for the spec and the spec could be banged out in less than a man month. The piece that is missing is the assertions and the conformance test for the ABI.  You could say, we’d do it if we had resources and move on.
Bill- you’re saying to take xkb wholesale and leave the creation of the conformance test as an exercise?
George- yes,  
Bill- I worry about that, it would be a big hair ball of a conformance test.  I’m not sure any existing platform would pass.
George- how many are augmenting live xkb?
Bill- They’re not augmenting, its just that the one or two existing xkb implementations, there are only a couple of x servers in common use, maybe 4. It is my impression that there are xkb bugs in all of them.  
George- that’s a good thing to know.
Bill- Its not bad a thing.  Its just that if we wrote a good conformance test, I’m not sure anyone would pass.  A conformance test is a prerequisite to people fixing their bugs.  My hesitation was taking the whole of xkb for the sake of convenience,  to avoid having to sub-specifying it.  It is appealing, it means we don’t have to do the work of triaging XKB, and saying which sections we care about. To some extent we already drafted that in the document we circulated early on.   The question is should we persue that document and in banging out a spec and through that document in. I don’t want to get into too much detail that may not be of interest to the whole group but I thought we should point out there is a missing piece in the keyboard group’s work and get input on how we should complete it.        
George- LSB has some library test- LSB libcheck, that can go through and open up the library and check for the existence of all the ABI’s and the runtime library.  They are in the process of writing, LSB appcheck which puts a layer between the application and the library.  It watches the parameters going in and out ABI to make sure they are in range.  The only thing it is missing is the behavioral part.  Testing the lib to see if you actually do something with it, system related 
Bill- because it’s client server arch it’s not synchronous, etc.
George- back in the 80's they did write the big X test suite, Open group did write it for Motif. It is possible to do those things if there are the resources.  
Bill- are you saying there will be an attempt to do that for applications in Xwindows within LSB/ FSG in the next couple of years?  If there is going to be an effort in that direction, we could say our conformance tests would come on line when we get this framework.  It sounds like FSG and LSG are moving toward more sophisticated conformance testing.  If that’s the case we could benefit from that.
George- Right now the LSB has out of the 4000 ABI’s defined has about 20% code coverage. Most of that was inherited from the Opengroup and a company Apptest that donated their old unix tests to the FSG.  There are about 30% to 40% more that they could get from them.  The company wanted a porting fee of about $250,000.  The LSB could get about 50% code coverage.  Then how do they get the other 50% coverage- write it or is there more stuff to buy.  They need to resource that.  Its not sexy code to write so it is hard to resource this since few volunteers.  Need to get more creative to get the money.  Hoping with OSDL doing their desktop work, they will collaborate with FSG and we’ll get some real dollars to do that. 
Bill- Since the world doesn’t expect 100% coverage for these ABIs., it gets us a little bit off the hook. Its not an unachievable obligation to bring in the whole ABI.
George- when someone does their conformance test with the FSG for accessibility, from ABI point of view, the distribution or the  application says I am following the ABI specification and the test helped to find problems for me but even if I pass or I don’t pass, I still follow the specification because the spec is the gold master.
Bill- okay just because you pass the test, that is evidence of your conformance but that isn’t the sum total of what you are asserting.  

George- some could say Linux distribution B, you failed this test.  It will go to the FSG specification authority, they will read the specification, they will go back and will look at test assertion and the test case and will make a ruling as to who is at fault- is it the test case, the spec or the distro.  Test case is wrong but spec is right, Distro B you’ll have to fix it.  But if B got it wrong, everyone else did too.  They’ll have to make an allowance unless B make some kind of aumentation.  
Bill- unless they regressed it, it is an undetected flaw
George- normally the only reason people diverge is if they are at a different release of something that added new features, they changed something or have patches other people don’t. The test cases help catch those divergences.  The specs are normally the gold master for everyone.
 If this group would go ahead and create an XKB ABI spec, Pull in the man pages, they can at least get people to say they follow this as opposed to something else or some augmentation of it. Then put the test cases on the wish list then ask FSG for funding ... library has 70 some ABI calls, this would take about a man year or two to write.  For OpenITN, I had IBM Yamato write some white character test cases, they assigned one guy and he did about 50 of them in one year. 
Bill- my recollection is the XKB ABI has more than 70 calls
George- Whitpro wrote all the PAM interface conformance tests and it took about 9 calendar months and donated it to the LSB. That’s just an example.  Libraries change over time, they may not be stable.  There can be several versions.  XKB shouldn’t change. But say we standardize on XKB 1.1 but you know 2.0 will be coming out.  Create a window where FSB accessibility 1.0 defines XKB 1,1.  For a two year window you are specifying 1.1, even though 2.0 is out but you put that stake in the ground to keep people at that stable level for a while.  Once FSG 2.0 is out maybe you’ll bump up to XKB 2.0.  
Bill- to take those as an example, if the new XKB rev’s were backward compatible everything would be fine.  If they broke compatibility, then 
Geoge- same would be true for live atk that the applications link into. It something were to augment to cause the runtime to be different, the workgroup could put a stake in the ground to keep desktop application stable.  At a latter time you could bump the rev.  What the LSB does is take all the libraries they define and produce LSB conforming header files and stub libraries and create a wrapper around GCC that builds against those header files to force the ABI to that level of the LSB.  So everyone is using   it to keep it complaint with that level.  Keeps applications at a back level lower set that you can repoduce over a period of time. There may be no forseeable changes in XKB. It Gives the workgroup to interact with LSB, the certification authority, Jim Zemlin and the board and show that we have produced a ABI, Specs and is included in certification.  Valuable as a marketing tool and promote accessibility at conferences..

Janina- that is very useful.  This sounds like it is meaningful follow on to the XKB. The next question is the SPI.  Some what competing activity since it is the same people.
Earl- sounds like FSB is not going to be impressed until we have the ABI.  Does that re-prioritize what we should do.  Does it make the xkb more important than AT-SPI from acceptance stand point.
George- not sure how FSG would make a product standard around the keyboard functional spec. AND get the distros to follow that.  Today, they use the ABI’s to say they conform to.  Not saying it can’t be done, just not sure how to do it functionally.    
Bill- the behavior conformance will have to be addressed eventually.
George- OSDL have brought everyone in the desktop to get things done.  LSB desktop spec will pull in live atk.  Seems to be happening in the desktop spec.

Bill- my interpretation of what Earl is saying, not a lot more work for the keyboard group to do the spec as an ABI spec.  If we chose to do it wholesale all we have to do is point to the man pages and the spec itself.
George- posted a couple of examples of ABI specs to the list with a table with footnotes that pointed to the spec. Just a couple of pages, prolog would be longer than the spec..  These are the terms.  The hardest thing is to take the ABI and import it into the LSB database so the spec is dynamically created.  Everything is dynamically built.
Bill- the LSB ABI engine would confirm that they header files are present, had all the symbols
George- effort is getting all that imported
Bill- need a stable place for the xkb spec., need a non-transient home on the web
George- LSB has seen that problem, they have a page called rough-specs.  The XKB spec could go there and we could reference.  Finish it off, hand it off to the FSG board.  Then work on ABI, that will be coming later. Keep things on track, get good interaction with the board.  Board will ask what is your road map.  
Bill- we made some extension to ATK as followup to activities of AT-SPI, do you see any conflict with the current ATK and version that has not been released?  
George- it should be okay, LSB    
Bill- as long as they get into the version soon, it will be okay?
George- go with what they have today, Gnome 2.14, those will be around for a couple of years.
If doesn’t make the 2.14 cut, that gives you time until next round of LSB.  LSB is now an ISO standard, once you get accessibility into the LSB,  it goes all the way to the international level.  It is getting more formal and mature.  

Janina- this has been a useful discussion, glad we have audio of it.  We need resume this conversation.  We were talking last week about when we could meet for a face to face.  Possible Bill would be available in mid July for developers meeting.   -
OSDL cooperation statement.         
Olaf- don’t know what is happening on the Gnome side.  I had the impression that not really taking off in the declaration. I haven’t been at meeting. If several projects come together to sign such a statement, it would be good to join that statement.
Peter- haven’t been a party to any of the more general Gnome discussions.
Bill- may be appropriate for me to add my own signature. Gnome is a much looser organization than KDE, there is no one person that can speak for it.  Several high profile people but nobody in charge.  There have been a few people but there is no hierarchy.  If you ask everyone, everyone looks around a waits.
Olaf- what is best way to proceed.  Gnome list
Bill- on Gnome webpage, there are few official projects that are active- internationalization.  Ask the people that are active, that are active in those projects.  I could ask on behalf of the Gnome accessibility project.  Could get someone from translation or international. Something like Gtk, why would they, KDE hasn’t shown interest in Gtk, doesn’t make sense, but on something like  Window manager there may be cooperation.
Olaf- one of the area for cooperation betwenn Gtk and KDE are freedesktop.org specs like icon themes.
Bill- Gtk people, see if they agree with that general statement.  Cooperation doesn’t mean KDE will use Gtk and Gtk will use qt, it means there are areas of common interest. Possible: Gnome translation project, Gnome Accesibility, gnome usability are the only active projects.  Gstreamer maybe
Olaf- they aren’t pure Gnome projects, gstreamer doesn’t understand itself as a Gnome project because they also want to be used by KDE 4.

Bill- they are not pure Gnome projects so I would think they would be happy to sign on. Approach them, perhaps the maintainers for GTK .  
Olaf- that is good idea, if we could add some of the Gnome projects, that would help.  
Bill- ask specific people.  One of the strengths of Gnome is that it is not hierarchical but it is a problem at times.  
Olaf- similar at KDE.  
Janina- Okay so Olaf will you be taking this as an action item?
Olaf- I will talk to Aaron and explain it.  
Janina- think about goals for this year and try to put some dates to these items.

John Goldthwaite
jgoldthwaite at yahoo.com
828 885-5304
Yahoo! Photos – Showcase holiday pictures in hardcover
 Photo Books. You design it and we’ll bind it!
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.linux-foundation.org/pipermail/accessibility/attachments/20060119/3f1622d2/attachment.htm

More information about the Accessibility mailing list