Re: [need for advices] fastCaching

Brent Cook wrote: > > On Mon, 22 Jul 2002 fabien.marty@free.fr wrote: > > > > A quick review of the code shows no handling of corrupt cache files. > > > I suggest writing the file length into the cache file and some > > > signature > > > bytes > > > as a prefix/suffix to the cache data to ensure data integrity. What about to have the md5 of the data as part of the file name? > > With FileLocking on (default), it's very very difficult to have > > corrupted cache files. But adding a test is possible. A md5 hash would > > be perfect but we would have to do some benchs. Because adding only file > > length has to be faster I think. > > > > Fabien > > I've had instances of corrupt cache on Digital Unix using NFS pretty > often (mostly empty files), so this is important, though it could be > optional. > - Brent > You may want to read the big Warning about flock() in the manual :-). Perhaps we could implement some application level lock ala Netscape (it may be nice for the installer too). Or well, md5() sounds good too. Tomas V.V.Cox