Re: Some ideas (was: bad ftp URLs make cache messy)

From: Kolics Bertold, University of Veszprem <bertold@dont-contact.us>
Date: Thu, 13 Mar 1997 00:05:28 +0100 (MET)

On Wed, 12 Mar 1997, Jan Klabacka wrote:
> Why squid can not check (and remember) file size on ftp server and
> after such interruption (or any erroneous interruption, or maybe
> even after all finished ftp transfers) just remove such incomplete
> file from cache (because it is anyway just rubbish) ?
Actually, there is size checking in the squid code, but I could not answer
why broken objects can still occur...

> Anyway, I am not using ftp cache at all (although it speeds up
> things), because of strange directory listings (not the worse
> problem), but also that it adds extension to files without them (for
> instance READMEs - so that with Netscape they can only be stored to
> disk). This is really strange.
This is not as simple - there was much discussion about that. Look at the
squid-users archive.

> Another idea (I am probably not using proper terms, but I guess it
> could be understandable): If error happens during fetch of object
> (mostly nonexistent) and squid reports error like requested url
> cannot be retrieved (I do not remember exactly), cache remembers
> this occasion for reasonably long time (configurable I guess).
It *is* configurable. If you tune it properly, it works fine.

> Yet another idea: anonymizing option in squid.1.1.8 is always
> removing type of browser. I know that it is some security risk to
> know type of client, but:
> 1. it breaks functionality of client sometimes (there are servers
> which behave differently with different clients)
I think that a standard-compliant (or oriented) web service should not
rely on that info, but on current HTML standards :-).

> 3. anonymizing - as far as I can see - has different reason than to
> hide client types (and thus break functionality). It should firstly
> hide site where request came from.
> I think it should be at least configurable which headers are to be
> removed during anonymizing (at least during compilation time -
> something similar like mime-types).
It is on the todo list. Look at the head part of src/http-anon.c

Cheers,
Bertold
==-==--==--==--==--==--==--==--==--==--==--==--==--==--==--==--==--==--==--==
 Kolics, Bertold E-Mail: bertold@tohotom.vein.hu
 University of Veszprem, Hungary W3: http://tohotom.vein.hu/~bertold/
 Information Engineering Course
==-==--==--==--==--==--==--==--==--==--==--==--==--==--==--==--==--==--==--==
Received on Wed Mar 12 1997 - 15:07:11 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:34:41 MST