[RFC] sha1 of fileid naming for knit files

[John Arbash Meinel]

Robert Collins wrote:
> On Tue, 2006-10-31 at 17:40 -0600, John Arbash Meinel wrote:
>> Anyway, mapping through sha1 would be okay, and it would mean we have
>> fixed size records, and don't have to worry about escaping the path on
>> disk. But we really need the reverse mapping first.
> 
> Because user data cannot control the path ids used, the sha1 approach
> seems ok to me as-is, though I too would prefer a reverse mapping in the
> index to allow detection of collisions.
> 
> -Rob
> 

I'm not as worried about collisions as I am about debugging. Because
without the backwards mapping, if there is a problem in one of the
files, you have no idea what text file it is associated with. I suppose
you could write a script to iterate over the inventory ids, and compute
the sha hash of each one.

But still, it is another layer of indirection, which can cause problems.

John
=:->

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 254 bytes
Desc: OpenPGP digital signature
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20061031/2368b3a9/attachment.pgp