On Tue, Dec 07, 2004 at 10:59:41PM +0100, Andreas Schuldei wrote: > this week (50) i attend the AFS hackathon in stockholm.
i participated in the hacking part of the events and prove-read some of the more interesting talks beforehand (and talked with the speakers, who were developers at the hackathon). OpenAFS is about to remove some long standing bottlenacks (like support for 2.6 kernels, ipv6, kerberos5, (and perhaps even ldap support!). disconnected (notebook) support is not beeing worked on, however. the windows driver is in better shape then the linux driver, currently. it allready now supports kerberos5, for instance. it seems to be very slick and has support for automatic (mass) installation on clients, even. on the downside the current version does not support windows versions before win2k. i tried to convice the developer to support even win98 (since many schools use it) but that is an entirely differnt architecture, with big gaping security holes and not the development efford would be significant (so it wont happen, unless it is paid for). i talked also with sam hartman, the debian maintainer of the MIT kerberos packages and also a kerberos and openafs developer. he was keen to help but has very little time. he will look for local debian people with afs know how to help out. my work on the openafs package to make user-mode-linux cross compiling possible will now also go in, after i fixed the user-mode-linux extraversion bug and packages can install cleanly. i myself spend the time writing some more modules for rootstrap (e.g. preseeding, tar-config, base, aux-install) and fixing more bugs there. i tried to push that upstream, but mdz is in spain and not responsive to mail or irc. furthermore i used a package that sam build some years ago as a template to do installation at install time. in the night from thursday to friday i helped fixing a srewed up ldap database. this week i will - try to send patches to sam integrating the initial krb5 and afs logic into packages - put uml machines demoing openafs and kerberos online - work on bofhd to actually allow to do something - compile jbofh into binaries, eliminating some dependencies - attempt again to build the cerebrum documentation Further future goals are to - shrink the import script in size (some more redundancy to remove). - work on the wlus testframework, so that work on wlus can be picked up again in a controlled way - write a cerebrum backend for wlus - switch wlus to that cerebrum backend
