Mailing List Archive

Arena bugfix still needed...
I just downloaded the latest Sun4 Arena version, and it still doesn't
grok "text/html;level=3" --- it downloads the file OK, but it displays
it as plaintext, NOT as HTML. Perhaps I should have checked myself
earlier, but we may not want to remove that particular workaround just
yet...

rst
Re: Arena bugfix still needed... [ In reply to ]
>
> I just downloaded the latest Sun4 Arena version, and it still doesn't
> grok "text/html;level=3" --- it downloads the file OK, but it displays
> it as plaintext, NOT as HTML. Perhaps I should have checked myself
> earlier, but we may not want to remove that particular workaround just
> yet...

drat.

Does anyone use Arena for anything other than testing ?
I noticed that it's broken in other places, so much so that
it's useless as a day-to-day browser.

If so, who gives a damn if perfectly reasonable HTTP headers
cause it to misfunction ? - I don't. As long as the browsers
people really use work ok, that's good enough for me.


robh
Re: Arena bugfix still needed... [ In reply to ]
/*
* "Re: Arena bugfix still needed..." by Rob Hartill <hartill@ooo.lanl.gov>
* written Fri, 5 May 95 13:42:41 MDT
*
** Actually, there isn't a single browser on the market which handles
** parameters on content types correctly --- they *all* wind up
** treating it as an unknown content-type, generally by dumping the
** file onto disk. This is not desirable.
*
* Netscape just worked ok with "Content-type: text/html;version=3" It
* may have guessed, but it did treat it as HTML
*/

Netscape 1.0 wasn't house trained in this area. Netscape 1.1 is
because we had to support

Content-type: text/html; charset=iso-8850-1

or similar. In fact, the 1.1 server will send parameters on
content-types but it has a gruesome hack in there to make it so it
will only return them to Netscape 1.1 or higher.

--Rob
Re: Arena bugfix still needed... [ In reply to ]
> Unfortunately, Arena is the *only* browser that people can currently
> use to demo content negotiation for HTML3. If there was another one,
> no matter what, I'd nuke the kludge myself in an instant.

How do they demonstrate content negotiation if they can't
handle the responses it can produce ???

This isn't our bug. Let the Arena folk fix it.

Hindering Apache content negotiation just so that Arean can
demonstrate it isn't a convincing arguement for me.

A recent usenet posting showed that people want to send parameters
after the content type. I think it's more important to cater
for legitimate use than a broken browser which people don't use.


robh
Re: Arena bugfix still needed... [ In reply to ]
> How do they demonstrate content negotiation if they can't
> handle the responses it can produce ???

> By talking to servers which produce responses it CAN handle, including
> Apache 0.6.2... granted it's their bug. Granted, they should fix it.
> Granted, they were asked to fix it more than a month ago. But right now,
> there is at least one browser I *can* use to demonstrate content negotation
> for handling HTML2->3 upwards compatibility, and I do not want to lose that
> until there's another.

but introducing a bug (which is what happens when you discard
valid parameters) to work around someone else's bug, stinks.

If the only reason to have this is so that *you* can demonstrate
content negotiation, then I can't see why *you* can't tweak your
own copy of Apache to misbehave accordingly. Put it in with
#ifdefs if you really need it (default = send params)

1.3 allowed parameters AFAIK. Apache should too.

robh
Re: Arena bugfix still needed... [ In reply to ]
> Actually, there isn't a single browser on the market which handles
> parameters on content types correctly --- they *all* wind up treating
> it as an unknown content-type, generally by dumping the file onto
> disk. This is not desirable.

Netscape just worked ok with "Content-type: text/html;version=3"
It may have guessed, but it did treat it as HTML

> 1.3 allowed parameters AFAIK. Apache should too.
>
> It didn't strip them off --- and if present, they broke every browser
> that anyone has. Why should that continue?

So as to encourage browser developers to fix their browsers ?
To allow people to use paramters to test clients ?

> A final point --- I gave this a +1 vote under the impression that you
> had in fact checked it with Arena,

I had, and it appeared to work. Not sure what I did then, 'cos
I can't repeat it now.

> and that it did in fact worked.
> Indeed, your CHANGES file says as much. Had I known that it broke
> Arena, I surely would have vetoed it.

Isn't there a way you can send back parameter-free responses ? (I don't know)

Anyway, I'll wait for someone else to comment and then go with the flow.

robh
Re: Arena bugfix still needed... [ In reply to ]
> I had, and it appeared to work. Not sure what I did then, 'cos
> I can't repeat it now.

Okay, it works with 0.93y which I ran from a SunOs machine,
but not from my 0.96s which I ran from a HPsUX machine.


now that's stupid.
Re: Arena bugfix still needed... [ In reply to ]
Are there any browsers which pass on parameter info to
external applications ? how about Netscape, Rob M ?

There was a call to have things like the BASE URL passed
to applications, when it appeared in the content-type header.
There's a .mailcap trick to specify any parameter name to
be used as an argument for the application, e.g. soemthing
like

application/postscript; pdfghostview %s -url [base]

where [base] is replaced with the value assigned to the parameter
"base" in the content type.

Does anyone have a .mailcap entry that uses parameter info ?

robh
Re: Arena bugfix still needed... [ In reply to ]
Unfortunately, Arena is the *only* browser that people can currently
use to demo content negotiation for HTML3. If there was another one,
no matter what, I'd nuke the kludge myself in an instant.

rst
Re: Arena bugfix still needed... [ In reply to ]
How do they demonstrate content negotiation if they can't
handle the responses it can produce ???

By talking to servers which produce responses it CAN handle, including
Apache 0.6.2... granted it's their bug. Granted, they should fix it.
Granted, they were asked to fix it more than a month ago. But right now,
there is at least one browser I *can* use to demonstrate content negotation
for handling HTML2->3 upwards compatibility, and I do not want to lose that
until there's another.

rst
Re: Arena bugfix still needed... [ In reply to ]
From: Rob Hartill <hartill@ooo.lanl.gov>
Date: Fri, 5 May 95 13:08:58 MDT

but introducing a bug (which is what happens when you discard
valid parameters) to work around someone else's bug, stinks.

Actually, there isn't a single browser on the market which handles
parameters on content types correctly --- they *all* wind up treating
it as an unknown content-type, generally by dumping the file onto
disk. This is not desirable.

If the only reason to have this is so that *you* can demonstrate
content negotiation, then I can't see why *you* can't tweak your
own copy of Apache to misbehave accordingly. Put it in with
#ifdefs if you really need it (default = send params)

It isn't just me. The only reason that ANYONE is interested in
content negotiation these days is for negotiating HTML feature sets,
and if we haven't got a browser that demos that, we might as well dike
out the code.

I don't object to private hacks --- in fact, I've got a few already,
for back compatibility with my old *.doit stuff. But I VIOLENTLY
object to this being one of them.

1.3 allowed parameters AFAIK. Apache should too.

It didn't strip them off --- and if present, they broke every browser
that anyone has. Why should that continue?

A final point --- I gave this a +1 vote under the impression that you
had in fact checked it with Arena, and that it did in fact worked.
Indeed, your CHANGES file says as much. Had I known that it broke
Arena, I surely would have vetoed it.

rst