Re: markup validator short-term roadmap

Hi Bjoern, all.

On Jun 8, 2006, at 14:24, Bjoern Hoehrmann wrote:

> HEAD has more bugs, less features, more dependencies, interesting only
> to people who would like `check` to have less code (and I think I was
> the only one who removed code from check, others would like it to have
> /more/ code, like additional SOAP templates) and to people who need the
> performance gains, which is us at most.

Thanks for your thoughts, Bjoern, but I think you're being modest, here.
* Less features, yes, I discussed that in my previous mail.
* More dependencies is not a big issue, is it? That can be documented, 
and except for opensp (and spo for the moment), all the dependencies 
are covered by cpan.
* Less (or more) code is not a goal per se. Code moves around, some is 
removed from check, but then s:p:o is quite a few lines of code. What I 
would think matters is that the code is maintainable, and I think what 
we have in HEAD is moving into that direction.
* More bugs, really? I haven't counted the HEAD-specific bugs, but I 
recall many instances of "this is fixed in HEAD", enough to make me 
doubt that statement.
* and finally, the performance gains... Maybe it interests us at most, 
but we're still our main customers, and we still have a  damning load 
problem on our two servers, so the performance gains, if they actually 
are as good as your tests showed, are a crucial change.

All considered, I wouldn't say it's an obvious winner, but it certainly 
deserves a bit more enthusiasm. Notably, if we consider the few 
scenarios for the future of the tool... either it finally evolves into 
something bigger, that can finally do XML decently, validate using 
schema, rng, etc - or it gets stripped down to a minima (s:p:o ?) and 
included in such a bigger tool, or it just stays at its current stage 
and gets replaced little by little with other, newer tools. Out of 
these 3 scenarios, I see 2 which would benefit from the s:p:o based 
architecture, and none really harmed by it.


>  Considering how previous alpha tests drove attention to the code, I 
> am not sure what the point would be. It doesn't seem to help convince 
> anyone to fix the known bugs.

That's a fair point, I agree that the alpha/beta tests have not always 
been particularly fruitful in bug fixes. They bring in some bug 
reports, but obviously not as many as just releasing into production.

I'm not sure what your stance is, though. Do you think it's useless to 
do an alpha test and we should release, instead, when we think it's 
time; or do you think the alpha test, and the release, etc, would be 
useless?

Thanks,
-- 
olivier

Received on Friday, 9 June 2006 01:04:19 UTC