From mboxrd@z Thu Jan 1 00:00:00 1970 Message-ID: <3D36E77C.9070102@nas.com> From: Jack Johnson User-Agent: Mozilla/5.0 (Windows; U; WinNT4.0; en-US; rv:1.0.0) Gecko/20020530 MIME-Version: 1.0 To: 9fans@cse.psu.edu Subject: Re: [9fans] cdrom floppy tape etc, media mount point References: <07032038068eb2473dd63c7a4e51d270@vitanuova.com> Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Date: Thu, 18 Jul 2002 09:06:20 -0700 Topicbox-Message-UUID: d0f7e330-eaca-11e9-9e20-41e7f4b1d025 >>thus, having done: >> >> autodir /n >> >>you can do, say: >> >> mount /srv/factotum /n/bletheridoo I was thinking about this last night on the drive home. Though I realize things will probably change with the new fileserver, I was thinking about how to get dump functionality from venti as-is. It occurred to me that one should be able to do something like: vac -svf $home/vac/`{date -n} -h ventiserver /n/hullabaloo run via cron, then have a variation on autodirfs that rebuilds the directory trees dump-style (or some custom style: 2002/Feb/20) based on the contents of $home/vac, and calls vacfs when traversing to the appropriate date. (tangent ahead -- leave while you can) My mind then started wandering to a coat-check style Web service where you'd post a file or set of files and vac would return a "ticket" so you could retrieve it later. Maybe your resume that you only use once a decade, or some file you'd like to share with friends that lies somewhere in between needing security and being public information. It also reminded me of the work done at PARC with small handheld devices called PARCTABs that have minimal storage, running much like a Plan 9 terminal. One thing I remember reading about them is that they were designed so people could treat them as if they had storage, so you could drag documents to the Tab, walk down the hall and copy them to another workstation, workboard, etc. In reality, the data never moved, but the representation of the data behaved as if it did, just like icons for files on a local hard drive. The part that makes me smile the most in Sean's paper is where he says, "For a user, it appears that vac compresses any amount of data down to 45 bytes." I think my wristwatch will hold 45 bytes. I can only imagine how many vac fingerprints a cell phone could hold. Where vac fingerprints are more powerful than PARCTABs, in my opinion, is that they're not tied to the network, per se. Fingerprints can be stored on flash media and popped in a pocket, written on a 3x5 index card and dropped in the mail, kept on a USB keychain, sent via pager, etc. With a vacfs port to Inferno and the Inferno IE plug-in (or vacfs support like WebDAV), you might be able to utilize vac URLs like vac://hullabaloo/64daefaecc4df4b5cb48a368b361ef56012a4f46. I think it would/will be fun to experiment with vac fingerprints as tickets for data using extremely small, cheap devices or Web interfaces or combinations of the two, along the same lines as Sean's idea of implementing SFSRO on top of venti. All of this, of course, for read-only data. I also wonder if there might be a convenient way to do chaffing and winnowing using a venti archive, given the amalgam of blocks. Sorry for the blather. I'm prone to it. -Jack