Re: New Branch Release

olivier Thereaux <ot@w3.org> wrote:

>Could you suggest what you would consider getting it out of the door without
>"rushing"?

Well, for one thing, not introducing it in a minor revision. :-)


>Would that work for you? I am not convinced it is necessary, but if it
>can alleviate your fears, I'm happy to accommodate.

How about making this a part of the normal beta cycle for 0.8 (possibly with a
somewhat extended beta period), and announce a preview of the API is available
on v.w.o:8001 to encourage discussion and feedback on the API?


>>I also have some reservations about the implementation, but I haven't
>>looked at it all that carefully yet.
>
>If joe luser sent a message to www-validator stating "soandso support is
>probably broken, but I'm not telling you why exactly, because I haven't
>looked at it carefully", they would be dismissed at not serious, don't
>you think? :) The feature has been on qa-dev for more than three months
>now: if you have a hunch that it may be wrong, please give it an actual
>look and help fix it, or find a good reason for trashing it.

Well, I meant to imply that, while I do have some reservations, these are based
on an incomplete study of the code in question so they may well be irrelevant.


But since you insist;

I'm skeptical of the implementation of an API in the form of an output template.
An API is an interactive thing where an output format is a one-shot, fire and
forget, operation. That's why the XML output is an output format while the early
proof of concept SOAP interface was done using SOAP::Lite and implemented as a
parallel code path to the CGI.pm one.

It also needs a client SDK and usage and regression tests; and it needs to be
tested with the common client frameworks.

I'm also not convinced the SOAP API should be a a part of the “check” CGI
script. The proof of concept demonstrated that this lead to several awkward
spots in the code, and quite a lot of bloat; which is one of the reasons why I
eventually nuked it.

This latter suggests this should wait for better m12n of the Validator code base
so the CGI and SOAP front ends can share the same back end libraries while still
being free to be implemented in the manner most suited to its function.


-- 
“It's not the mere technical details of inserting the live round into the
 chamber, pointing the weapon at one's foot, and pulling the trigger, but
 rather, it's about the advisability of doing that in the first place.”
                                             -- Alan J. Flavell on ciwah

Received on Monday, 6 February 2006 08:45:57 UTC