Yahoo admits mangling e-mail

Yahoo has admitted that it has been changing some of the words inside attachments accompanying messages sent to users of its web-based e-mail system.

The company said it had taken the action to prevent the spread of viruses.

However, critics have said that the filtering system is too crude and is having unforeseen effects.

Foolish filter

Some users of the Yahoo.com mail service are complaining that the changed words are scrambling the sense of academic papers and business documents.

Yahoo's filtering system was intended to stop the spread of viruses that lurk inside the computer code that defines the layout and appearance of a webpage.

This code, called HyperText Markup Language (HTML), can be made more interactive, animated and lively using another language called Javascript.

Yahoo word changes
Eval to review
Mocha to espresso
Expression to statement
Javascript to java-script
Jscript to j-script
Vbscript to vb-script
Livescript to live-script

When web browsers encounter a chunk of Javascript, they automatically try to do what the code demands.

The code is behind such things as visitor counters and can change the appearance of buttons and bars on a page when a mouse moves over them.

Unfortunately, some computer virus writers also use Javascript to set off malicious programs that, among other things, can kidnap personal information such as passwords or credit card numbers.

To prevent this happening Yahoo has been scanning HTML attachments on e-mail messages sent to Yahoo.com users looking for common commands that can be subverted.

The practice has been going on since at least March 2001.

Scrambled words

Many people have found that Yahoo's filtering system is too crude and is scrambling innocent words.

Unfortunately, the filter does not respect the boundaries of words. This means that when banned text appears inside another word, it gets converted too.

This has led to the appearance of strange words such as "medireview" instead of "medieval" and "reviewuate" instead of "evaluate".

The filtering has scrambled e-mail addresses and led to puzzling changes in academic papers.

Search engine Google lists over 600 sites using "medireview".

One article in the online edition of the Hindu newspaper even talks about the "medireview" Mughal emperors of India seemingly unaware that the word has no proper meaning.

Yahoo has yet to say whether it plans to stop checking and changing attachments.