Coda File System

Re: Coda Progress

From: Brian Bartholomew <>
Date: Thu, 4 Dec 1997 15:10:54 -0500
Bob knows what I want, and like him I am interested in Mail Handler
files, ical calendars, web and C source code.  Let me explain the
architecture I want:

The Coda server sits in an ISP co-location rack, where it has 10 Mbits
or better to the Internet.  It accepts incoming mail for,
serves web for, and serves  It has a scsi DAT
tape jukebox, and backs up its local disk to local tape nightly.  I
can telnet into the Coda server as a shell account when I have no
computers with me and I need to access my stuff.

One of the Coda clients is on my laptop.  I write source code on the
laptop, inc mail and maintain my calendar.  These modified files
eventually percolate to the Coda server, which backs them up.  Some of
the source code I modify is html in the web server documents
directory, which eventually is served as live content.

Another Coda client is on my desktop machine at home.  I use it just
like the laptop, and it has only a modem or ISDN-speed connection.
Given sufficient connectivity the Coda server and all clients hold a
consistant set of files.  Given insufficient connectivity the out of
date files are still useful.  I never have to remember what file I'm
working on that need to be copied where, or which copy is newer, as
the system tends to correctness.  The laptop disk is much smaller than
the Coda server disk farm, yet I still have access to all the server
files under identical pathnames.  Coda pages files as necessary.

With suitable system administration arrangement the long-term stable
laptop and desktop client disk contents are completely disposable.
All user data is eventually flushed to the Coda server and backed up.
The Linux system content is either backed up or recreatable in an
automated manner.

The next improvement is to have multiple geographically dispersed
co-located Coda servers which reconcile changes with each other.  This
keeps any single ISP connectivity or server disk failure from taking
me out.  I suspect the right number of servers is four, because that
will tolerate a single failure and still protect me with three
machines to vote on changes.

Another member of the League for Programming Freedom (LPF)
Brian Bartholomew - - - Working Version, Cambridge, MA
Received on 1997-12-04 15:26:45