Re: Require UAs and intermediary caches to assume Vary: User-Agent

In message <>, Willy Tarreau writes:
>On Wed, Oct 17, 2012 at 08:15:51AM +0000, Poul-Henning Kamp wrote:

>> Fine with me, provided you _first_ reduce the number of distinct
>> U-A strings on the web to something reasonable, lets say no more than 25 ?
>One of my customers encodes them in logs to reduce log size, so basically
>each new UA string gets assigned a 16-bit ID which is reused everytime the
>same string is encountered. Last time we checked, we saw something like
>30k different strings in one day worth of logs ! Some browsers emit all
>the combinations of their plugins and versions, making the UA currently

Many of the, now illegal in EU, tracking schemes use U-A as a major
component in disambiguation factor, for exactly this reason.

>Another benefit of getting much less different UAs is that their
>identification string could significantly shrink then.

Personally I think the U-A is just plain wrong and should be dropped.

Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.

Received on Thursday, 18 October 2012 07:27:36 UTC