Re: [Afpfs-ng-devel] UAMs

[Alex d. <ale...@gm...>]

Derrik,

Sounds great!  I'd love to see your code.  Doing this alone is getting
a bit tiring.

I have actually rewritten the UAMs interface with the expectation of
writing new UAM support, but I haven't written any new ones.  As you
mentioned, clearly the old interface was lacking.

So as to facilitate developement, I've just pumped out 0.3.3 which you
can download.  You'll be able to see my UAMs interface. Now, if this
isn't a sufficiently flexible interface, I have no problems whatsoever
in changing it.  Just send me a patch.

And a note about 0.3.3... this is a release which is much better than
0.3.2, but it has some known problems, such as:
- leaking memory on writes (I know the problem now)
- on concurrent reads and writes, there are some problems where the
accelerated dsi packet processing gets out of sync.  I've solved about
10 of these problems in the last week.
- I know it has problems with fuse 2.6.1 (with threading, a problem in
fuse I think), and is untested against fuse 2.5.3.  Works with 2.6.0.

But, I have succesfully been able to copy a kernel tarball, untar it
and compile it with a loopback connection to netatalk.  I can get part
of the way with a mount to an OSX machine.

I'd really prefer that people hang on until version 0.4 before using
afpfs-ng, although I'm at the point where quality is converging (ie.
I'm fixing bugs at a faster rate than I'm finding them).


- Alex

On 1/30/07, Derrik Pates <de...@de...> wrote:
> All,
>
> I've actually progressively been writing my own AFP library, as well as a
> FUSE-based tool to mount an AFP (over TCP) filesystem on Linux - however,
> I wrote mine completely in Perl. I'd thought to start adapting my UAM
> code to afpfs-ng - I have implemented UAMs for both 1- and 2-way random
> number auth, as well as DHCAST128 and DHX2 UAMs.
>
> Unfortunately the existing interface for UAMs is not sufficient for
> anything other than plain-text authentication (which is, by the way,
> limited to 8-character passwords). Having a function that just generates
> an authinfo block isn't enough for UAMs like DHX2, where 'FPLogin' is
> called, followed by two different invocations of 'FPLoginCont'. I can
> provide my implementations as examples, but the preferred way would be to
> have one function that accepts the username and password (and AFP version,
> if needed - it might already be indicated in the AFP session object).
>
> I've begun reading the libgcrypt API documentation, which has the necessary
> interfaces for DES and CAST5 encryption, CBC, and MD5 hashing, but
> without retooling the UAM interface, it won't be possible to practically
> implement the more advanced UAMs.
>
> --
> Derrik Pates
> de...@de...
>
> -------------------------------------------------------------------------
> Take Surveys. Earn Cash. Influence the Future of IT
> Join SourceForge.net's Techsay panel and you'll get the chance to share your
> opinions on IT & business topics through brief surveys - and earn cash
> http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
> _______________________________________________
> Afpfs-ng-devel mailing list
> Afp...@li...
> https://lists.sourceforge.net/lists/listinfo/afpfs-ng-devel
>