🚀 go-pugleaf

DISCLAIMER: I am not a developer of the HURD, I just peek at the snapshots
            from time to time.  All answers subject to sudden net.correction.

In article <3lntur$vnl@lynx.dac.neu.edu>,
zachary brown <zbrown@lynx.dac.neu.edu> wrote:
>I know that unix was not developed with security in mind, and so is weak in
>that area. Since the HURD is not unix, I was wondering how it will handle
>security, and will it be better than traditional unix systems? What are the
>issues of HURD security? What will be its soft spots? What about sendmail?

In the Hurd, there is a daemon named auth which handles authorizations.  Each
authorization is a list of UIDs and GIDs.  Every process gets, from its parent
or from the password daemon, an auth port identifying who it is allowed to
act as.  When the process wants to use a service, it and the server both
contact auth, which passes the client's UIDs/GIDs to the server. 

It's pretty slick: a user can provide a virtual file system but can't lie and
serve out setuid files owned by joe bloe since the daemon can't make an
authorization by joe bloe.  Even more intriguing is the thought of you and joe
running your own auth server so that your servers will recognize each other's
IDs.  This won't affect other users, tho, since they will all be using the
system's auth.

The basic security issues are the same, although this scheme provides a much
more flexible framework. (i.e. using UIDs like VMS Identifiers, etc)  If you
read an unlimited line into a local fixed size buffer, it doesn't matter
what OS you are running.  Unless they rewrite sendmail (not likely unless
copyright needs dictate), it will have the same strengths and weaknesses.
Don't forget, the Hurd's designed to look like Unix to the user.  
"Security-Challenged" is a feature.  :-)

	Eric B.