Re: new feature: parameter for specific User-Agent identification

olivier Thereaux wrote:

>> Sounds good, I hope "forward" covers "no UA"
[...]
>> Maybe you need to map an empty "other string" to "no UA".
[...]

> I hadn't thought much about "no UA". To me it is usually
> not a good idea to not have a UA header at all - that's
> the prerogative of broken bots - should we really have
> that?

UA is a SHOULD in HTTP, so there might be valid reasons to
have no UA header field at all.  But I've no idea what an
*empty* UA field should do, the syntax doesn't permit this.

> This makes me think maybe when a UA is requested we should
> use that custom of adding a (Compatible; W3C Markup 
> Validator <version)) ?

Certainly allowed by the syntax.  But I thought the idea is 
to check what the server does with the "real" UA.  A server
looking for say "mozilla/3.0 (OS/2 U)" or similar could be
confused by embellishments added by the validator.  Only a
wild guess.  The concept of server behaviour depending on
the UA string makes me nervous, unless it's about desperate
attempts to identify broken UAs.

 Frank

Received on Thursday, 19 June 2008 05:55:30 UTC