Getting 3.0 out (Was: Offer to update the build documentation ...)

[Boris Kolpackov <bo...@codesynthesis.com>]

Hi Scott,

Scott Cantor <ca...@osu.edu> writes:

> What's left to do? I know some platforms don't work, but honestly
> either the project supports those platforms or it doesn't.

Well, for starters, 3.0 does not build on my stock Debian box.
This appears to be due to ICU being installed on my machine
and autoconf picking it up and configuring Xerces-C++ to use
it (which is not a bad idea by itself). But, as further
investigation showed, autoconf support for ICU is not
finished. I tried to change a few things but ran into more
problems after regenerating the configure scripts with newer
version of libtool. See the following thread for more information:

http://thread.gmane.org/gmane.text.xml.xerces-c.devel/8275

If anybody has any idea how to fix this, please let us know.


> It's not reasonable to me to hold up development for a platform or compiler
> for which nobody is donating time, a box to test on, or both.

I agree, if there is no way to test something then there is no way
to support it. However, as I mentioned above, there are problems
on mainstream platforms or across platforms.


> Put up or shut up basically.

Remember that there was a build system which, while being ugly and all,
supported quite a few platforms. Now we threw that build system away
and replaced it with another one which does not (yet) work properly
on less mainstream platforms. It seems to me that telling users of
those platform to fix it themselves or shut up is a sure way to
alienate them.

Boris

-- 
Boris Kolpackov
Code Synthesis Tools CC
http://www.codesynthesis.com
Open-Source, Cross-Platform C++ XML Data Binding

---------------------------------------------------------------------
To unsubscribe, e-mail: c-dev-unsubscribe@xerces.apache.org
For additional commands, e-mail: c-dev-help@xerces.apache.org

RE: Getting 3.0 out (Was: Offer to update the build documentation ...)

[Scott Cantor <ca...@osu.edu>]

> I can guarantee you that if that is going to be the attitude of the
> development team, for what are very popular platforms (mostly, any
> platform with ICU), then this project will likely lose a large portion
> of its' user base.

I'm not on the development team. I'm just an upstream consumer who believes
that 3.0 should not be beholden to exception cases. I suspect that the
definition of exception case is "whatever I don't need", the same as any
other project. If ICU is "very popular" it's certainly news to me, but if
that's a dealbreaker than we should track it as such, shouldn't we?

> If Xerces wants to stop claiming to support ICU, I
> suppose that's a way out of this as well, but that will certainly
> drive me to some other solution.

I know nothing about ICU, but I'll repeat my earlier statement: is this
problem in Jira? If not, I have little or no sympathy. If so, people with
ICU expertise should be looking at it; if we have no such contributors, then
how on earth can we support ICU? Isn't that common sense?
 
> As it is, the new Intel XML
> libraries are looking awfully tempting for win32 and x86-linux, given
> their performance and support, and the 3.0 release discussion
> certainly isn't convincing me to stay with Xerces.

We have to be realistic. I don't see how pretending we have the ability to
support things we can't support helps anybody. The project isn't static;
something that was supportable may become unsupportable if the resources
simply aren't there anymore.

I'll repeat: poor support is much worse than none. If the project can't
deliver what you need, it helps nobody to pretend that it can.

-- Scott



---------------------------------------------------------------------
To unsubscribe, e-mail: c-dev-unsubscribe@xerces.apache.org
For additional commands, e-mail: c-dev-help@xerces.apache.org

Re: Getting 3.0 out (Was: Offer to update the build documentation ...)

[Alberto Massari <am...@datadirect.com>]

James Berry wrote:
> Hi Nicholas,
>
> On Nov 5, 2007, at 12:44 PM, Nicholas Bastin wrote:
>
>> Scott Cantor <ca...@osu.edu> writes:
>>> Put up or shut up basically.
>>
>> I can guarantee you that if that is going to be the attitude of the 
>> development team, for what are very popular platforms (mostly, any 
>> platform with ICU), then this project will likely lose a large 
>> portion of its' user base.  If Xerces wants to stop claiming to 
>> support ICU, I suppose that's a way out of this as well, but that 
>> will certainly drive me to some other solution.  As it is, the new 
>> Intel XML libraries are looking awfully tempting for win32 and 
>> x86-linux, given their performance and support, and the 3.0 release 
>> discussion certainly isn't convincing me to stay with Xerces.
>
> It certainly is not the intent of the development team to not support 
> icu. It looks to me like the icu build was broken by some additions 
> for an icu message loader (Alby?). That clearly needs to be fixed.
>
The first error reported by Boris was that the message loaders using 
additional files (ICU and MsgCat) implemented the 'install' target 
through a rule that was imported from Makefile.incl, and wasn't 
rewritten (and this has been fixed in August). As for the newer libtool 
not putting the dependency in the samples command line, I couldn't see 
it as I was using an old Ubuntu distribution; now that I upgraded to the 
latest one, I could try reproducing it. But having someone else working 
on autoconf would be great, as I am mainly a Windows developer with some 
Linux/Solaris experience.

Alberto

P.S. Someone should step up to fix the xerces-p binding too, or my next 
work in that area will be to tell autoconf not to generate/build 
anything in the swig subfolder

---------------------------------------------------------------------
To unsubscribe, e-mail: c-dev-unsubscribe@xerces.apache.org
For additional commands, e-mail: c-dev-help@xerces.apache.org

Re: Getting 3.0 out (Was: Offer to update the build documentation ...)

[Nicholas Bastin <nb...@opnet.com>]

Scott Cantor <ca...@osu.edu> writes:
> Put up or shut up basically.

I can guarantee you that if that is going to be the attitude of the  
development team, for what are very popular platforms (mostly, any  
platform with ICU), then this project will likely lose a large portion  
of its' user base.  If Xerces wants to stop claiming to support ICU, I  
suppose that's a way out of this as well, but that will certainly  
drive me to some other solution.  As it is, the new Intel XML  
libraries are looking awfully tempting for win32 and x86-linux, given  
their performance and support, and the 3.0 release discussion  
certainly isn't convincing me to stay with Xerces.

--
Nick

---------------------------------------------------------------------
To unsubscribe, e-mail: c-dev-unsubscribe@xerces.apache.org
For additional commands, e-mail: c-dev-help@xerces.apache.org

RE: Getting 3.0 out (Was: Offer to update the build documentation ...)

[Scott Cantor <ca...@osu.edu>]

> Well, for starters, 3.0 does not build on my stock Debian box.
> This appears to be due to ICU being installed on my machine
> and autoconf picking it up and configuring Xerces-C++ to use
> it (which is not a bad idea by itself). But, as further
> investigation showed, autoconf support for ICU is not
> finished. I tried to change a few things but ran into more
> problems after regenerating the configure scripts with newer
> version of libtool. See the following thread for more information:
> 
> http://thread.gmane.org/gmane.text.xml.xerces-c.devel/8275
> 
> If anybody has any idea how to fix this, please let us know.

I'll certainly take a look. I haven't ever used ICU, but I might be able to
identify some things.

> I agree, if there is no way to test something then there is no way
> to support it. However, as I mentioned above, there are problems
> on mainstream platforms or across platforms.

We need an accurate list of issues, I think. Is this one in Jira?

> Remember that there was a build system which, while being ugly and all,
> supported quite a few platforms. Now we threw that build system away
> and replaced it with another one which does not (yet) work properly
> on less mainstream platforms. It seems to me that telling users of
> those platform to fix it themselves or shut up is a sure way to
> alienate them.

Projects also have to be realistic when they have constrained resources.
Poor and non-ongoing support *is* worse than none, and if the autoconf
support is there, it's much easier to fix the outliers than it was before.
Losing platforms in the short run will pay off in the long run.

I don't see how the answer can be anything other than fix it yourself if
there's nobody available with the platform to do it for them. It's not being
difficult, it's just the way it is. Obviously we can get access to Debian,
that's an example where we can fix it, but HP-UX or AIX might not be.
 
-- Scott



---------------------------------------------------------------------
To unsubscribe, e-mail: c-dev-unsubscribe@xerces.apache.org
For additional commands, e-mail: c-dev-help@xerces.apache.org