On Fri, 9 Nov 2001, Lex Spoon wrote:

> 
> Bijan Parsia <[EMAIL PROTECTED]> wrote:
[snip]
> > True, but so? :) Detecting user-agent is something of a hack, but it can
> > be a very useful one. Personally, I'd like to have that for use with
> > *normal* swikis, though content negotiation is prolly the right way to do
> > this.
> 
> In what case would you use this, exactly?

I would use it to send raw swiki text, for example, to Scamper. Or to
serve Squeak object based pages to those browsers that understand
them. This is why I say "content negotiation is prolly the right...".

[snip]
> Most of the time people propose a feature like this, they say "I want to
> send tag X if it's browser B" or "I want to send text-friendly HTML if
> it's browser B".

Nope. I want to send stuff altogether different :) I may well want to send
*much less*.

>  Well, HTML already provides for both of these
> problems.  Browsers are pretty good at just ignoring tags they do not
> support.  And keep in mind, there are more text-mode browsers than just
> lynx, and there are more Squeak browsers than just Scamper. 

Yes?

> Furthermore, your guesses about what someone will desire based on their
> browser, have very good chances to be mistaken.  It would really bug me,
> for example, if someone were to cut the IMG tags out of a page because
> they realize I am using lynx.

That would be absurd, given alt-attributes. How about *tables* though :)

I'm really talking about HTML being the *least* common denominator.

> Alternatively, someone might want to layout the page completely
> differently depending on the client.  However, this is a losing battle
> -- new browser versions come out all the time.

Not really. Netscape 4.75 and IE 5.5 seem pretty stable choices at the
moment.

Note: I *totally* agree that using standards based (X)HTML etc. is,
generically, a good thing. However, discriminating by user-agent for some
purposes can be useful. This hardly seems wildly controversial.

> > And it's not *quite* true that you have to modify the sever for new
> > browsers...if the browsers forge their user-agent. Abuse, yes, but often a
> > useful one ;)
> 
> First, not all browsers do this.

I'm surprised. Examples?

>  Should people really have to
> reconfigure their browser to visit your site?

Often have to anyway.

>  Second, this isn't going
> to achieve the desired result.  If an Opera user visits your site, do
> you really want to send them a random choice between Netscape and IE
> HTML, depending on what their User-Agent happens to be set to?

Well, again, it seems we have *very* different purposes in mind. I'm not
talking about munging a bit of HTML to optimize for a certain browser. I'm
talking about entirely different functionality.

> Now, I will admit that if you are happy with supporting 60-80% of the
> people who visit your site, then you can just support the top 3-4
> browser versions in their most common configurations.  However, you can
> usually support 100% of the people who visit your site if you stick with
> standard HTML.

"Standard" being the big issue. If you support MS browsers, standards are
hard to support. Vice versa. :)

User-agent discrimination is a hack, often a useful one, especially for
experimental stuff and often when you control both the server and the
client. As we do with ComSwiki + Scamper. There are other ways to achieve
the kinds of effects I'm talking about, of course, and often they should
be prefered. Not always.

Cheers,
Bijan Parsia.

Reply via email to