On Sat, 2005-02-12 at 15:01 +0100, Gottfried Szing wrote: > hi > > >> I am a big fan of moin, and it does all of these. It is simple, > >> easy to hack on, has a good security record, and generally rocks. > >> MediaWiki is also a good choice. > > > > > > I just install moin and the README has a task that says to warn when > > upgrading because the wiki does not always work properly when > > upgraded. This is a serious concern, data preservation is all > > important. > > to jeff: this was one of the reasons why we have decided to use a wiki > which uses a db as backend.
I am going to put forward some general concepts, not to necessarily to deny what you are saying, just to give reasons why this is not a global truth. Unfortunately using a DB does not guarantee that a program will not corrupt / loose your data on upgrade. The information may be there but suddenly not accessible because of a subtle change of logic. Indeed due to the additional complexity of using a database I would consider this more likely. > another reason was that IMHO it is easier to move the db to another > hardware (separation of frontend and backend) or a new server. if the > server that is containing the data runs out of space, moving all the > data to a new server and just changing the db-connection settings is > easier to do than moving a whole filesystem and mounting via NFS or SMB > or sth else. This is a valid reason however you forgot some details: - Network latency - how much impact would having a network in the middle have on the overall impact on throughput. - There is about 10% overhead in using a database over a simple file in most cases. Databases can have an impact on performance where their indexing facilities subset the data more efficiently. The problem domain of a Wiki maps very well onto a filesystem. > btw: i wouldnt use a filesystem mounted from a remote server as > datastorage with excesive data access (like a wiki usually does). Nor would I. I might duplicate using rsync across that link for a warm / hot swap. > and IMHO using a DB as backend is not a bad idea in the first place: > AFAIK wikipedia.org uses also a DB as a backend and there it seems to work. There may be some justification for a huge number of pages of a public diverse website. The wiki may end up in a single directory and the directory access alone may be significant. Does not make it right for every application, the problem domain is important. > cu, gottfried -- Ken Foskey OpenOffice.org developer -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
