Blog of Leonid Mamchenkov

If you ever mention that your web application uses database to store files, you risk being flamed into oblivion.  Indeed, in most cases, it is a bad idea, since file system is more effecient when it comes to files.  However, there are cases when it makes sense to have files saved in the database.

Maybe I am doing something wrong, but in the last six month, I had to develop at least three systems that used MySQL for file storage (uploaded files that have to be synchronized across several hosts, etc).  Yesterday, for the third time I stumbled across the same problem, that almost drove me insane.

MySQL has four data types for storing binary data - TINYBLOB, BLOB, MEDIUMBLOB, and LONGBLOB.  Somehow I always forget about these and use BLOB.  BLOB works just fine, but it has a limit on size, which is rather low - 64 KBytes.  The mean thing here is that it will work just fine with most of the test data - text files, short PDFs, and small pictures.  Once the application is tested and put into production, the corrupted files will start coming in.  Re-writing all parts that deal with uploading, moving, cleaning, escaping, and encrypting binary data takes time.  Going through file reading and writing routines is boring too, and it won’t help either.

By the time, the issue is discovered and all fields are changed to LONGBLOB, it is often very late, and you’ve lost your weekend, as well as a lot of large files. This post is an attempt to save my (and your) sanity.

Reminder: use LONGBLOB instead of BLOB for file storage, unless you are absolutely sure about the maximum size of incoming data.

Slashdot is running the post about annoying software.  The fact that Slashdot crowd mostly consists of computer geeks is sort of a guarantee for some interesting comments.

With my Fedora 9 saga I had to review and try a lot of new software.  Needless to say, I found quite a few annoying bits.  Here is a brief list, just to give you an idea:

• Clock applet in Gnome. It shows calendar with Sunday being first day of the week.  If you don’t like it, you’ll have to recompile your locale to change it. This one is cancelled out though by an excellent support of Google Calendar (or, for that matter, any other web published calendar).
• Metacity window manager in Gnome. Window titles are displayed in the middle.  This is really annoying for those of us who are used to seeing them on the left.  There is no option to change this setting either in GUI or in GConf.
• Pidgin new message notification. I once had it popping up nice looking bubbles, but I don’t remember how I managed to do it.  I also don’t remember how I managed to break it.  And I have no idea to bring them back.  I really miss them though.
• WordPress 2.5 post editing screen. It has been much reworked in the latest version and looks and feels so much better. However, the list of categories was moved from a really convenient location on the right of the screen to a really inconvenient location at the bottom of the screen.
• FileZilla FTP manager. This one drives me nuts with server connections.  It either disconnects every 40 seconds when being idle.  Or it keeps multiple connections open forever and most FTP servers block me out temporary.
• Request Tracker (RT3). Works perfectly with queues and tickets, but annoys the heck out of me when I need to do something with users.  Users aren’t first level citizens, like tickets.
• SugarCRM. Excellent business tool, with lots of small annoyances, like not being able to set default user role, disable theme selector everywhere, change logos to company ones, lock down the functionality, etc.  Most of these are easily fixable.  But some aren’t as trivial as they may sound or seem.
• Google Reader. This one annoys me a bit (but often) when I want to leave a few items in the feed unread and go deeper into archives.  Somehow it keeps marking everything I passed as read.

Now, what piece of software were you annoyed with recently?

There was an sudden boost in searches for Cyprus yesterday.  The reason for that was a bug in Twinkle - a Twitter client for iPhone.  For some reason, Twinkle was identifying current location wrong for a whole lot of users.  Instead of being somewhere in the States, they were said to be Nicosia, Cyprus.  That probably felt very surprising for them, especially considering the fact that not many of them knew what or where Nicosia, Cyprus was.  Hence, all the searching.

There is a rumor going around that it was the work of Cyprus government, desperately trying to keep tourism levels up.  You, of course, should believe whatever your tin foil hat tells you to believe…

Anyone who had ever wrote more than 3 lines of code will tell you any time that all software has bugs. That’s just the way it is.

And while I don’t need any reminders of this fact (mainly due to me writing a lot of code at any given week), I got one special today.

A SPAM comment was posted to this blog, although you haven’t seen it because it went to moderation, that was clearly a result of a bug in SPAM software. The message contained a long list of phrases like ‘Thank you’, ‘Very interesting’, and ‘I bookmarked your blog’. Obviously these are intended for link SPAM. But they were supposed to be used one at a time. Oops.

Yet another bug that I submitted to Red Hat Bugzilla got fixed. This time it took more than a year - from 21 May 2004 until 12 September 2005 - but a workaround was found much earlier and could be easily applied.

P.S.: I was stupid enough not to notice that the bug was closed and add another comment today. Nuthead!