🚀 go-pugleaf

kitchin@intersurf.com writes:

>For virtual domain web hosting my ISP just switched over to shared IP's
>via HTTP_HOST (see below if you don't know what this is).   I complained,
>since I have only seen this used on very low cost providers, like icom.com and
>webcom.com.  But my ISP claims the following:

>> In accordance with InterNIC, ARIN, MCI and UUnet policies, we have
>> eliminated the use of IP's for virtual web sites permanently.
>>
>> Browsers supporting only HTTPD v1.0 will indeed see the wrong web
>> site, for the reason you cited above.  Updated browsers have been
>> available for far more than one year supporting HTTP_HOST.
>>

Most sites now use features that make them incompatible with HTTP v1.0,
anyway, so it would seem to me that this is the right way to go.

If the server does put in the page of pointers, it means one extra
mouse click to get to the page they wanted.

Probably the pointer page would also have a link to where to download a more
modern browser.

-- glen

In <6f6jvo$a78$1@nntp6.u.washington.edu> gah@u.washington.edu (G. Herrmannsfeldt) writes:

>kitchin@intersurf.com writes:

>>For virtual domain web hosting my ISP just switched over to shared IP's
>>via HTTP_HOST (see below if you don't know what this is).   I complained,

No, they are via the Host: header.  HTTP_HOST is simply a CGI 
environment variable that is set.  

>>since I have only seen this used on very low cost providers, like icom.com and
>>webcom.com.  But my ISP claims the following:

>>> In accordance with InterNIC, ARIN, MCI and UUnet policies, we have
>>> eliminated the use of IP's for virtual web sites permanently.
>>>
>>> Browsers supporting only HTTPD v1.0 will indeed see the wrong web

Erm... someone doesn't know what they are talking about.  "HTTPD" would
be a HTTP daemon or something, eg. "NCSA httpd".  It is HTTP/1.0.
And most clients, with the exception of IE4, make HTTP/1.0 requests.
Most still send the Host: header, since just because it is defined
in HTTP/1.1 and not in 1.0 doesn't mean that 1.0 clients can't send it.

>>> site, for the reason you cited above.  Updated browsers have been
>>> available for far more than one year supporting HTTP_HOST.
>>>

>Most sites now use features that make them incompatible with HTTP v1.0,
>anyway, so it would seem to me that this is the right way to go.

Erm... don't think so.  If you have anything to support this wild
claim, please post it.

As I said earlier, the only common client that makes HTTP/1.1 requests
is IE4.

In ba.internet G. Herrmannsfeldt <gah@u.washington.edu> wrote:
> Most sites now use features that make them incompatible with HTTP v1.0,
> anyway, so it would seem to me that this is the right way to go.

How so? Besides host-header-based virtualhosts, what other
post-HTTP/1.0 features are commonly required?

<sigh> I can just imagine sites adorned with "Best viewed with
HTTP/1.1 Keep-Alives" buttons.

> If the server does put in the page of pointers, it means one extra
> mouse click to get to the page they wanted.

Another problem with HTTP/1.0 based browsers accessing
host-header-based virtualhosts: all non-relative links will be
unusable.

_____________________________________________________________________
Anirvan Chatterjee  . anirvan @ chatterjee.net .  PGP 2048/0xE2D13BA9
Fast & free new/used multi-bookstore searching @ http://www.mxbf.com/

   It should be noted that HTTP/1.1 *requires* the Host: header... you can't
   send an HTTP/1.1 request without it or the WWW server will return an 
   error.

   This doesn't prevent HTTP/1.0 from sending a Host: header, and in fact
   most 1.0 browsers *do* send it.  Unfortunately, there are still a 
   significant number that do not, too many to tell them to go take 
   a running leap.

   Frankly, I think HTTP/1.1 is a good thing, and not just because it 
   requires the Host: header.  It fixes the persistent connection / proxy
   caching problem that HTTP/1.0 requests have, and that's major.  HTTP/1.0
   proxies can't do persistent connections reliably.  It's disappointing
   that browsers are slow to move to the new standard (which isn't so new
   at this point!).

					-Matt

-- 
    Matthew Dillon   Engineering, BEST Internet Communications, Inc.
    <dillon@best.net>, include original article w/ any response.
    do not under any circumstances send email to joeblow@bigspender.idiom.com
    and, for gods sake, don't email bucktram@popserver.idiom.com

In article <6f7pj6$678$1@flea.best.net>,  dillon@best.net (Matt Dillon) wrote:
>    proxies can't do persistent connections reliably.  It's disappointing
>    that browsers are slow to move to the new standard (which isn't so new
>    at this point!).

While there are a lot of new computers, there are also a lot of old
computers out there.  When I'm at home, I read mail and news on
a 16 Mhz 68030.  Netscape 2.0 barely runs on it...

Sam

-----== Posted via Deja News, The Leader in Internet Discussion ==-----
http://www.dejanews.com/   Now offering spam-free web-based newsreading

Anirvan Chatterjee <c..bojswbo@13.usenet.us.com> writes:

>In ba.internet G. Herrmannsfeldt <gah@u.washington.edu> wrote:
>> Most sites now use features that make them incompatible with HTTP v1.0,
>> anyway, so it would seem to me that this is the right way to go.

>How so? Besides host-header-based virtualhosts, what other
>post-HTTP/1.0 features are commonly required?

OK, they aren't post HTTP 1.0 features, but they are features that older
browsers don't have.  I just didn't understand the numbering.
A post in another group claims (I didn't verify it) that Netscape 2.x and
IE 3.x and up send the host header.  So, what features didn't exist before
Netscape 2.x and IE 3.x?  Well, I have tried a frame-less browser, and many
sites don't allow that.  They just say sorry, and supply links to download
a new one.

><sigh> I can just imagine sites adorned with "Best viewed with
>HTTP/1.1 Keep-Alives" buttons.

>> If the server does put in the page of pointers, it means one extra
>> mouse click to get to the page they wanted.

>Another problem with HTTP/1.0 based browsers accessing
>host-header-based virtualhosts: all non-relative links will be
>unusable.

Why would they be unusable?  They are links to the wrong host?

OK, you make a base host that displays a page with links to all
the virtual hosts, to be displayed by clients without the host header.
In this page you also include links to download a newer browser.
(You can even check the header, so you only recommend a new version of
the same browser.)  

You can make absolute links point to the base host, or any virtual host.
Relative links should not point to a different virtual host, but absolute
links should always work.

-- glen

In <6f8s9a$aoe$1@nnrp1.dejanews.com> spaik@simagic.com writes:

>In article <6f7pj6$678$1@flea.best.net>,  dillon@best.net (Matt Dillon) wrote:
>>    proxies can't do persistent connections reliably.  It's disappointing
>>    that browsers are slow to move to the new standard (which isn't so new
>>    at this point!).

>While there are a lot of new computers, there are also a lot of old
>computers out there.  When I'm at home, I read mail and news on
>a 16 Mhz 68030.  Netscape 2.0 barely runs on it...

Huh?  Because some people run old versions of clients, for whatever
reason, means that new versions of clients can't implement HTTP/1.1?

I have trouble following that.

albaugh@agames.com (Mike Albaugh) writes:

>	I suppose that in a world that accepts ActiveX, I shouldn't
>be surprised that some folks will do exactly that, but to take the
>approach "Hey, anybody who visits my web page had better damn well
>be prepared to turn his/her computer over to me" seems extreme, IMHO...

It makes a good filter.  Chances are good that people
like you (and me) don't want what these people are
offering anyway.

--kyler