PHP Classes

Hello

Recommend this page to a friend!

      Conteg Content Negotiation  >  All threads  >  Hello  >  (Un) Subscribe thread alerts  
Subject:Hello
Summary:Opening this Forum on Conteg, and saying 'Hello'
Messages:5
Author:Alex Kemp
Date:2007-03-10 13:46:58
Update:2007-10-08 07:17:57
 

  1. Hello   Reply   Report abuse  
Picture of Alex Kemp Alex Kemp - 2007-03-10 13:46:58
This is a message from the Class author, saying 'Hello'.

Bug reports welcome!
--------------------

Conteg is in active use on my and other sites, so it *does* get tested daily. That does not mean that it (or me) is perfect - far from it.

Please include:
1 Conteg version
2 Web-server name + version
3 Server OS + version
4 Browser name + version
5 Client OS + version

...and give as concise a report as you can of the problem.

You are also advised to obtain either Mozilla and/or Firefox and install the 'Live HTTP Headers' extension, which will display both Request and Response headers for each browser access (invaluable for developers).


Note: my email address is bogus
-------------------------------

Unfortunately, this site exposes the author email address. I've lost email access on 2 domains in the past due to this practice, and will not allow it to happen again (spam email to the bogus address is already in full flood).

I can be contacted by post or PM via the forums on my site.

I trust that Conteg will prove invaluable for you.


- Alex Kemp
modem-help.com/
forums.modem-help.com/

  2. Addition   Reply   Report abuse  
Picture of Nikita U. Nikita U. - 2007-06-04 18:55:55 - In reply to message 1 from Alex Kemp
When checking the 'Content-Encoding' header you exclude Netscape 4.x from browsers which are to be served compressed content. There's another browser you should think of handling in a special manner: MSIE 4, MSIE 5, MSIE 6 on windows without service packs. I'm not as mad as you might just have thought, here's what you should definitely read: http://lists.over.net/pipermail/mod_gzip/2002-December/006826.html
In short, since MSIE4 there was a problem with caching when 'Vary' header contains anything apart from 'User-Agent'. The bug was fixed in one of the service packs for IE6 under win xp (not sure which), but prior that even 'Expires' header didn't help.
Of course there aren't many windows without service packs, but still there are even less Netscape 4.x's :)

  3. Re: Hello   Reply   Report abuse  
Picture of Alex Kemp Alex Kemp - 2007-10-06 22:40:43 - In reply to message 2 from Nikita U.
Thank you for your response, Nikita, and I am sorry to leave it so long to reply (I set notifications for New-Thread messages only).

At this point I've read through the original message only. That issue *may* explain why one of my users is having a problem (http://forums.modem-help.co.uk/viewtopic.php?t=902) (MSIE 6.0 on Windows 2000), even though his issue is the opposite: page being displayed from cache instead of shown new.

Your message is certainly exactly of the kind that I wanted to receive, so I do thank you for it. It looks like some coding needs adding to take account of this issue. Typical MS.

I shall respond when I've completed my reading through that thread. Thanks again.

- Alex Kemp

  4. Re: Hello   Reply   Report abuse  
Picture of Alex Kemp Alex Kemp - 2007-10-07 09:36:28 - In reply to message 2 from Nikita U.
I've gotten through both that thread, and also the one that inspired it. Phew, whole lotta reading. And very useful. I still have the W3C.ORG thread to read through.

The following is now clear:

- 1 The handling of "Vary:" is currently wrong within Conteg.

The 'Vary' Response Header is sometimes included, sometimes not. That is not correct. If the 'Accept-Encoding' Request Header is (potentially) used to vary the response for a page, and 'Vary' is sent at all, then "Vary: Accept-Encoding" should be sent with that page on every occasion, whether the page is encoded or not. There should also be a way to unilaterally switch use of the 'Vary' header off, for those that choose not to use it.

- 2 MSIE 4.x response needs considering.

I've had 47 hits so far in October from v4.x versions (from 219,063 for all MSIE) (possibly is just two people!).

It is clear that (non-patched versions of) this browser cannot handle a "Vary: Accept-Encoding" response, and put up a false error message. Not sending the header, however, would compromise every other browser that passes through the same proxy. I believe that to be unacceptable.

- 3 MSIE 5.x and 6.x response needs considering.

MSIE 6 is 130,551 hits (46.3 %), and MSIE 5.x is 3,891 (1.4 %). Since MSIE 7 use is only 84,520 (30 %), that makes MSIE6 use the strongest on my site (Firefox is 17.9 %).

MSIE 5 + 6 can accept a "Vary: Accept-Encoding" header; if received, they treat it as if "Vary: *" was sent. The side-effect of that is that the browser does not cache the page on the computer. That means that re-visiting the same page causes a full refresh of the page (s/b a 304 from the server).

Losing the bandwidth savings from 304s is a blow, but since the typical Content-Encoding reduction is 76% (a four-fold reduction), I think that it is worth it. If the webmaster has the choice to switch 'Vary' off, then they can make their own mind up.

Fortunately, use of Conteg does not affect delivery of static files from the server at all, so the impetus for the `lists.over.net' thread does not apply here.

I have some testing to do on my server separate from this issue. That will take a couple of hours. I'll then code the changes into Conteg, and release it onto my site, plus report it here.

Many thanks for your message, once again, Nikita.

  5. Re: Hello   Reply   Report abuse  
Picture of Alex Kemp Alex Kemp - 2007-10-08 07:17:57 - In reply to message 2 from Nikita U.
OK. Conteg is now fully compliant in it's use of the Vary header. It is also possible to completely switch use of that header off.

The combination of these two should satisfy all the issues that you put forward.

The above is implemented in Conteg v0.13.5; I'm making a separate post on that.

Thanks again, Nikita!