Dirmngr overhauled

[Simon Josefsson]

Werner Koch <wk at gnupg.org> writes:

> On Fri, 12 Dec 2003 10:04:59 +0100, Werner Koch said:
>
>> I have not changed that becuase I recall that there used to be a lot
>> of trouble.  Given that I shifted my development from Sid to Woody
>> plus some newer tools I have not tested DB4. 
>
> Okay. I removed DB2 et al. in favor of a CDB implementaion.  Frankly
> that is what we have should since teh beginning: We do only create the
> cache database once and the tehre will be only read access.  When
> downloading a new CRL the cache DB is created froms cratch.  The CDB
> implementation is pretty straightforward and should work on all modern
> OSes (it mmaps its files).  Tehre is no need to do a dirmngr --flush
> because the code will itself detect the updated database file layout.

Thanks for your work, I'll get to testing it soon.

> One missing thing is some file locking code to avoid duplicate
> downloads of CRLs in case more than one process is using the cache.
> It should still work, but it might lead to a inefficient bandwidth
> usage.

There is TDB as well, which permit concurrent write access via a
DB-like API.  It seems nice, but I haven't tested it myself yet.