On 10/12/06 8:26 AM, "Chris Hostetter" <[EMAIL PROTECTED]> wrote:
> I agree ... comments various people have made over the last few weeks or > so have gotten me thinking that standardizing the way errors are > returned to the HTTP client (for both updates and selects) is one of the > biggest "API problems" in SOlr right now. This is a complicated issue. We sent zillions of messages about it on the Atom Publishing Protocol list. The end result was to use HTTP request methods and response codes and be religiously correct about the HTTP behavior. We should take a look at the APP spec to see what we can learn from that. Atom got *lots* of review, all the way to Roy Fielding. The current APP draft is here. This is in IETF Last Call, so it is nearly finished. http://bitworking.org/projects/atom/draft-ietf-atompub-protocol-11.html One option would be to implement APP rather than designing something new. When I was at HP, one of the design principles was "standard is better than better." APP will not be an exact match to Solr, but it is very well designed and it is probably worth the effort to build on the vast amount of work and review that went into APP. If you want to dig in further, the mailing list is here: http://www.imc.org/atom-protocol/index.html If we have APP for indexing, results in Atom format would help a lot. Google GData is a pretty good Atom search API. wunder -- Walter Underwood Search Guru, Netflix
