On 11/06/2012 01:22 PM, Junchang(Jason) Wang wrote:
Last night, Richard and me together compared your script (simply
returned E_SYNTAX) and ours (worked correctly) line by line for at least
10 minutes, and finally figured out there's a extra 's' in your script
which inured the problem.
Jason: Right, I discovered that later as well.
I wonder whether it's reasonable that servers return error code
(say E_SYNTAX) PLUS some clues (say, a piece of the string that incurs
the problem). I think that benefits clients a lot.
That is always a good option, but the best place to put this would
have been in definition of ErrorResourceEntity (c.f. Section 6.5.2
of the ALTO protocol document). I think it is too late to add it
now, but more importantly, I don't think we absolutely need this
functionality. HTTP is rich enough to provide for richer error
responses without us introducing mechanisms in ALTO.
As Richard mentions, a more appropriate HTTP status code to return
in this particular case would have been a 404, not 400. It is
the R-URI that was the problem here.
But there are other mechanisms to inform the client of the nature of the
error without necessarily nailing it down in the ALTO protocol
specification. A couple of that come to mind are:
1) Use a more descriptive Reason-Phrase. For instance, consider Test-
JSON-ERR-1 from draft-gurbani-alto-interop-cases-02. In that test
case, the server returns a "400 Bad Request". There is nothing to
stop the server returning "400 Mismatched braces in entity-body".
The Reason-Phrase is there for consumption of humans, so the more
descriptive it is, the better.
2) Use multi-part MIME. One part is the normal
"application/alto-error+json" and the other part could be
"text/plain" that describes the error in more detail.
Thanks,
- vijay
--
Vijay K. Gurbani, Bell Laboratories, Alcatel-Lucent
1960 Lucent Lane, Rm. 9C-533, Naperville, Illinois 60563 (USA)
Email: vkg@{bell-labs.com,acm.org} / [email protected]
Web: http://ect.bell-labs.com/who/vkg/
_______________________________________________
alto mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/alto
