[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: "Flaw" revisited (was: Bug? FSFS revision control)

From: Ben Collins-Sussman <sussman_at_collab.net>
Date: 2005-01-27 19:07:34 CET

[cc'ing to users@]

On Jan 27, 2005, at 8:09 AM, Dassi, Nasser wrote:
> Perhaps; when I usually develop software I personally incorporate
> various self-awareness and self-healing mechanisms to increase the
> confidence that the working environment is safe enough for the
> applications to operate predictably. 

That sort of self-checking stuff *is* throughout Subversion: that is,
svn is constantly verifying that data being passed around is "sane".
If you change a byte in a binary diff stream, I svn will notice and
barf. If you tweak a value in a BDB table to point to something
nonsensical, svn will notice the malformed structure and barf. There's
all sorts of validation going on. There are even checksums being
read/written/verified every time files are read or written to the
repository (and also when sent over the network.)

But it's impossible to detect *every* possible data corruption. The
fact is that some corruptions look like legitimate data -- they're
legitimately formed. And therefore those things won't cause problems
until way further down the line.

If you have a specific suggestion about how to detect the exact sort of
corruption that you caused, and if your suggestion doesn't take a
ridiculous amount of work, then please suggest it to the dev@ list. As
someone mentioned already, it's only worth "so much" effort to detect
corruption; if you can't trust the OS to safely store and retrieve
data, the whole foundation of any database is shot.

To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Thu Jan 27 23:13:43 2005

This is an archived mail posted to the Subversion Users mailing list.