Re: [squid-users] Re: Add a prefix/suffix if a domain is not resolved?

From: Amos Jeffries <squid3_at_treenet.co.nz>
Date: Mon, 07 Nov 2011 14:39:37 +1300

 On Sun, 6 Nov 2011 08:47:04 +0000 (UTC), Adam Gaskins wrote:
> Amos Jeffries,
>
> While I truly can appreciate your border-line pathological idealism,
> I think it's reasonable to desire the browser behaves as designed
> when behind a proxy. For better or worse, people have grown use to
> certain tools and features within their browser of choice. I do work
> in the IT field, but imagine how much benefit the average Jane/Joe
> can get from something like TDL guessing! Sorry to rant, but I am
> also seeking an answer to this. Do you have any information on
> actually making this work? There is almost certainly a work-around
> for this issue, I just haven't found it yet either.
>
> -Adam

 I don't quite follow your argument. For an analogy, people grown used
 to using forks are encouraged to continue to use forks even when eating
 soup. Different jobs == different "best" tool, and being accustomed to
 something is no measure of best.

 As for this feature proposal and Squid. Follow my logic...

 So, (a) do many (a few million) DNS lookups for potential variations of
 the name?
  up to 1.6 days wait for a single page to load while Squid scans for
 potential URLs. The why of it is outlined in my earlier post. I guess
 this is the same reason why browsers don't do it this way. Yes the
 timings will usually be shorter, but even the less than 60 second delay
 with Squid today trying to locate IPv6 domain access is raising a lot of
 complaints.

 Browsers avoid that major problem by using a search engine to do the
 domain lookup instead of DNS.

 So, that brings us to (b): which corporate search engine do we force
 everybody to use to follow the IE behaviour?
  Serious question. None of the Squid dev team believe we have the right
 to decide that on behalf of the whole Internet community.

 Browsers get away with it because they can (and do) provide users with
 a selection of engines to use. With the user always fully able to change
 the engine choice. We can build in configuration and provide it to the
 admin, but that still forces a whole network of users onto one search
 engine. Not much better.

 Squid does provide plugin interfaces for third-party scripts to do
 anything they like to fill a certain operation (auth, URL redirect, ACL
 test, file erasure, SSL cert creation).

 One option is (c) using the redirector interface with a helper that
 does all the lookups and searching.
  As Kinkie posted earlier in the thread. That could be done with a
 local database of who the users are and what their preferences are.
 Combined with a URL redirector that does all the preference loading and
 domain searching. Or not, if you want to be mean and not give the users
 any choice.
  Either way a simple "no change" or the "proper" URL to redirect the
 user to is all that is relevant to Squid.

 Or, (d) you could try to convince the browser people to do this name
 search before contacting the proxy as well as before doing a direct
 website connection. The PAC approach favoured by Henrik earlier goes a
 way towards that without needing to wait for the browser people to add a
 new feature.

 In summary, we have a choice between (a) very annoying behaviour, (b)
 very nasty behaviour, or leaving Squid unchanged (c and d), with a
 third-party script do all the tricky work to make everybody happy. We
 opt to keep Squid simple whenever possible.

 You still get to choose from (c) and (d), and best of all, you get to
 select through them how much user choice is followed and what name
 variations you want to allow.

 Amos
Received on Mon Nov 07 2011 - 01:39:42 MST

This archive was generated by hypermail 2.2.0 : Mon Nov 07 2011 - 12:00:02 MST