On Fri, 3 May 2002, Shachar Shemesh wrote:

>  From what I understood about this bug, it was irrelevant for Mozilla
> RC-1 because the exploited feature was not working at all there (i.e. -
> the bug was there, but another bug prevented exploiting the first one).
>
> I am using debian sid, and they carry nightly mozilla builds (mine is
> from 20020502, as can be seen from the "user-agent" string in this email).
>
> As far as I understood from the bugtraq posting, they tried to contact
> someone at mozilla (rather than submit the bug to bugzilla). Thisis
> standard responsible full disclosure procedure, which mozilla was
> probably not set up to follow (the procedure was built with commercial
> companies in mind, not open transparent development processes and bug
> systems).
>
> As a side note - I am begining to lose hope with securing a platform. I
> used to be able to say certain distributions make frequent security
> updates, and that this is the only real protection. However, I since
> amended this saying to add "and you need to support their automatic
> update to not halt your services". This new addition automatically
> disqualifies Microsoft (they have, on several occasions, posted updates
> that broke functionality, and they also posted updates that were not
> included in the automatic update mechanism). I have also had an occasion
> in the past where RedHat's up2date mechanism completely broke
> functionality of a desktop machine, so I am a bit hesitant to trust
> RedHat on this (anyone can give counter experience?).
>
> I am coming more and more to the conclusion that the only way is to
> forego functionality, and even that doesn't always save you (what good
> is saying "don't use telnet, use SSH" if ssh has had exploitable bufer
> overruns itself?).
>
> I'll start a comparison table, and I'm asking the list to help me fill
> it in:
>
>
>     Microsoft (W2K and Win XP server):
>
> *Unattended update mechanism*: I don't know (is "Windows Update"
> scriptable?)
> *Packages are protected against DNS poisening*: Yes
> *Update mechanism reliability reputation*: Very bad - patches take time
> to be included even after they are released. Machines regularily stop
> working after updates.
> *Responsivness to problems*: Mediocre
>
> Please note that responsivness and reliability are a tradeoff, but MS is
> doing poorly on both.
>
>
>     *OpenBSD*:
>
> *Unattended update mechanism*: I don't know for sure, but I think not.
> *Packages are protected against DNS poisening*: As far as I know updates
> are distributed source only, so the answer is probably not.
> *Update mechanism reliability reputation*: Very good, but I don't have
> many sources on this.
> *Responsivness to problems*: Usually high
>
>
>     RedHat:
>
> *Unattended update mechanism*: up2date
> *Packages are protected against DNS poisening*: RPMs areopionally
> signed, but I don't think up2date can be configured to reject unsigned
> RPMs. I don't even know whether it verifies the signature.

IIRC this is the default behaviour.

> *Update mechanism reliability reputation*: Bad personal experience, but
> not in a server. Anyone using it?
> *Responsivness to problems*: ?
>

up2date requires registration (unlike redhat itself). Does this pose any
technical problems?


>
>     Debian:
>
> Debian's table seem to be version dependant. There are fluctuations
> between sid, woody and potato.

(or between stable, which only has bug fixes and security fixes, to
'testing/frozen' that has changes that are usually a bit tried-and-tested,
to 'unstable' that is rumored to be stable enough?)

> *Unattended update mechanism*: apt-get. It has the extra capability of
> adding third parties to theupdate mechanism. They can also upgrade
> major versions (i.e. - from potato to woody). On the down side, the
> output from the upgrade scripts is not easy to automatically script
> (though possible).
> *Packages are protected against DNS poisening*: debs are signed using
> gpg. There is a package containing the signatures of all debian
> developers. However, apt does not have an option to verify the
> signatures. A script workaround may be possible, but the default answer
> is "no".
> *Update mechanism reliability reputation*: Very high. Debian maintainers
> on the stable release are extremely careful not to break anything. Apt
> on stable will not upgrade you to a new version if security problems
> have been discovered in the old one, but will rather install a patched
> older version that is not vulnerable. This causes false positives when
> scanning for vulnerabilities, but means your server doesn't break
> functionality or upgrades unexpectably.
> *Responsivness to problems*: Usually high, with at least one recorded
> glitch.
>
>
>     Sun:
>
> *Unattended update mechanism*: Not that I'm aware of.
> *Packages are protected against DNS poisening*: No. Regular HTTP download
> *Update mechanism reliability reputation*: Never heard of problems.
> *Responsivness to problems*: ?
>

Mandrake:

Anybody got their urpmi working properly?
Is there any way to use it for automatic periodical updates?

I personally upgrade packages using rpm after I gave up on
urpmi/rpmdrek/whatever . Has it improved in 8.2 ?

Assuming it works properly:

*Unattended update mechanism*: Probably not
*Packages are protected against DNS poisening*: Yes, standard RPM
  signatures
*Update mechanism reliability reputation*: Older versions occasionally
  don't get enough testing (e..g: apache/php packages for mandrake 7.2)
*Responsivness to problems*: ?

-- 
Tzafrir Cohen
mailto:[EMAIL PROTECTED]
http://www.technion.ac.il/~tzafrir



=================================================================
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word "unsubscribe" in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]

Reply via email to