-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 01/03/13 12:27, Jason Dictos wrote: > Overall SQLite has been the perfect solution to our situation.
Thanks for the details. What programming language do you use? (Dropbox uses Python.) > Early on we figured out quite quickly not leverage the read/write > locking models of SQLite Do you use WAL mode? I'm also curious roughly what you do with your schema. It looks like the Dropbox on Android schema has elements of implementing a tree with everything in one big table and no triggers. Do you use any of SQLite's extensibilty such as adding your own collations or user defined functions. > As it stands now, our customers periodically have random I/O errors > ... As you have large deployments, random stuff does happen. We have a web service with several requests per second from browsers all over the world and sadly have to use HTTP (cough *IE* cough) rather than the SSL we normally use. Even though TCP/IP is checksummed there are sporadic corruptions that come through (typically bit flips here and there). 8 years ago there were lots of CPU random errors: http://blogs.msdn.com/b/oldnewthing/archive/2005/04/12/407562.aspx Do you run an integrity check at startup on the database? I did so with BitPim, but we didn't have analytics so there was no idea how often corruption happened. There has been an open feature request for a while to have data checksums to deal with the case that what SQLite thinks it wrote is not what is later returned: http://www.sqlite.org/src/tktview?name=72b01a982a Roger -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.11 (GNU/Linux) iEYEARECAAYFAlExZhYACgkQmOOfHg372QTLBgCgsbidy6oQfmAeS4OWq4OBSmFI zxEAn04lneghgvr+ww76AQWzycZ3x+Q0 =eya6 -----END PGP SIGNATURE----- _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users