There is also great merit in maintaining an index file. You can index, say, the date, title and filename, and the index file will probably remain small enough that you *can* slurp it. I mean, you're not going to run Slashdot with it but even if you post once a day for the next three years you're still looking at a few K. And you can do a few neat tricks like just reading off the top three records to save time.

For great speeedness, use fixed-length records. Space wastage sucks, but you then get the luxury of being able to arbitrarily jump around the file.

<lecture props='grep'>
The thing is, once you you've got a good index file system going, you probably want to add querying abilities, so you'll add in a keyword index. Then you'll break the index handler off into a daemon process so you can update asynchronously and handle collisions. Then you'll generalise the data storage to handle arbitrary data in arbitrary table. You might even add a SQL interpreter. You get the idea - you will thank yourself for learning to use a database and DBI.
</lecture>

But if it's impossible, I meant what I said about an index file.