- From: Richard Cyganiak <richard@cyganiak.de>
- Date: Tue, 4 Sep 2012 22:32:27 +0100
- To: Lee Feigenbaum <lee@thefigtrees.net>
- Cc: public-rdf-dawg-comments@w3.org
Hi Lee, On 4 Sep 2012, at 21:24, Lee Feigenbaum wrote: >> I now believe that the best compromise between simplicity, interoperability and deployed practice is to encourage the use of text/plain error messages, perhaps as a SHOULD, or even just as a non-normatively stated possibility; the minimal option would be to encourage text/plain error messages by providing *only* text/plain examples (currently there's one for text/plain and one for HTML). Bonus points for encouraging a style where the first line of the text/plain message can be used as a one-line summary of the error. > > We're planning to non-normatively suggest (as the examples show) that the status line of the HTTP response give this plan text one-line summary of the error. We think that this is inline with a proper use of HTTP. What do you think? I'm not so keen on this approach. Custom HTTP status lines are hard to generate, and hard or even impossible to retrieve in HTTP clients. Quite a few SPARQL stores are sitting behind reverse proxies, which can do unpredictable things to your HTTP message. Keeping it simple might be best. Richard
Received on Tuesday, 4 September 2012 21:32:51 UTC