Hello Bob!
Generally I think that we should avoid implementing or including functions
that are already part of the jre distributions. I have had serious thoughts
on removing log4j in favor of the new (in jre1.4?) logging api. If this was
the only consideration alternatives 3) and 4) are the way forward.
There are other considerations, as you point out. Specifically the time
constraint.
I suggest that we set the ambition to alternative 3). If we, when we get
closer to the next stable release, have not yet managed to solve this, we
could make a quick fix to implement alternative 1.
/Linus
2012/1/2 Bob Tarling <[email protected]>
> We have a historical issue with the 1.3 to 1.4 conversion giving errors on
> large files
>
> Most recently issue 6383 has been reported (
> http://argouml.tigris.org/issues/show_bug.cgi?id=6383)
>
> We have previous issues 3992, 4188, 5188
>
> I've discovered replacing Suns default xalan processor with an old version
> of Saxon resolves this problem.
>
> How should we move forward on this?
>
> Options I see.
>
> 3) Fix the stylesheets so that they work with xalan
>
> The obvious favourite choice but I think if we understood the issue we
> would have done this already.
>
> 1) Distribute Saxon B and continue to use this for XSLT processing
>
> Simplest option.
>
> Disadvantage is lack of support. Saxon has dropped some functionality that
> we are using from their latest releases. This is from their website.
>
> Saxon-HE does not offer all the capabilities that were present in Saxon-B.
> Most notably, support for Saxon extension functions and other extensions
> was dropped, as was the capability for writing extension functions that
> rely on dynamic loading of Java or .NET code (a new facility for
> "integrated extension functions" is however available). Users whose code
> relies on these features of Saxon-B should either purchase the Professional
> Edition product or stick with Saxon-B: the latest release of Saxon-B is
> 9.1.0.8, and although there are no plans to develop it further or maintain
> it, it will remain available indefinitely.
>
> 2) Fix the stylesheets so that they are compatible with Saxon HE and
> distribute that.
>
> At least we are now fully supported but we have more work to do in some
> complex stylesheets and still need to distribute Saxon. I'm not sure if the
> xslt extensions used can be replaced easily with standard xslt.
>
> 4) Drop the UML1.3 to UML1.4 conversion from the standard ArgoUML
> distribution and move it to an optional module.
>
> If a user tries to load a 1.3 file then they would be prompted to download
> and install this module.
>
> Advantage - smaller footprint for core ArgoUML with Saxon only in the
> module download.
>
> Disadvantage - another module to maintain.
>
>
> Any thoughts....?
>
------------------------------------------------------
http://argouml.tigris.org/ds/viewMessage.do?dsForumId=450&dsMessageId=2903708
To unsubscribe from this discussion, e-mail:
[[email protected]].
To be allowed to post to the list contact the mailing list moderator, email:
[[email protected]]
