Greg Thomas wrote:
> On 09 May 2005 11:12:10 -0500, kfogel@collab.net wrote:
>
>> Don't blame Berkely DB, blame Subversion's, shall we say, unusual way
>> of using Berkeley DB.
>>
>> (I just feel bad that Berkeley DB and Sleepycat software are getting
>> undeserved bad press because of Subversion's idiosyncracies in how it
>> uses BDB.)
>
> Although given that svn is using BDB in a legitimate manner (albeit
> unusual) and BDB is failing to cope with it, this would suggest a flaw
> in BDB - though I recognise that Sleepycat are endeavouring to fix it.
Actually, the issue is that Subversion does not comply with the requirements
given on one specific page of the BDB programmers' reference manual.
The only flaw in BDB is that it leaves itself open to corruption unless
certain external guidelines are not followed, and that the manual does not
emphasize this vitally important point above the rest of the text.
As a result, it took assistance from Sleepycat to work out what the problem
was, but now they are going to integrate the support for the kind of failure
detection that any BDB-using application of Subversion's general structure
needs into BDB 4.4.
Max.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Tue May 10 12:29:47 2005