[XML-SIG] xml vs. sql

Martin v. Loewis martin@v.loewis.de
Tue, 8 Jan 2002 00:03:25 +0100


> I'm considering building a data-centric website using xml instead of the
> usual SQL database. I expect the site will receive about 1000 hits per
> day. There will never be more than a couple thousand records in the
> data. I'm planning on using Apache, mod_python and Python's minidom. 

If you are looking into a Python implementation, I recommend to use
pickle/cPickle over XML. If desired, you can still offer XML
import/export.

> The benefits of doing without the database are:
>     1. the data will be highly portable

If, by that, you mean "across operating systems and software
versions", then this will be about pickles as well.

>     2. the data will not be subject to corruption

Depends on how you operate on them. It is certainly possible to
truncate or otherwise garble XML files.

>     3. no need for DB admin

The same would be true for pickles.

> Potential downsides:
>     1. Slow processing under load

pickles likely load much faster than XML is being parsed.

> Any thing else I need to consider? 

Complexity of the source code. An XML application will be many more
lines than using pickles.

> Anyone care to share past experiences doing the same?

mailman uses pickles for its databases, and I believe many other tools
do as well. 4Suite offers pickle support for all their DOM
implementations.

Regards,
Martin