Quote:
Originally Posted by HarryT
Undoubtedly, but it's a lot harder to make devices handle errors "gracefully" than it is to handle correct code, because there are a virtually limitless variety of ways in which things can go wrong.
|
Some years ago (I don't remember when) a programmer from Poland (that's why I remember the event) annoyed with web browsers that crashed randomly on many pages made a tool for generating random HTML pages, which were nearly correct (ie. they were generated correct, and then the program randomly made one of two changes in each generated page, like switching places of two closing tags, removing a closing tag, adding a random opening or closing tag in random place, adding a randomly generated bytes in random place in the file. And made this tool public with a blog post (or something similar). It got in the news, and shortly after that all the browsers rather quickly stopped crashing on any such pages.
Perhaps if there existed such a tool for generating nearly correct ePub documents, with faults and unexpected data in as many places as possible, more device/software manufacturers would test their viewers more extensively with it? As it is, they probably write a small set of tests, or test the documents on a few random pages, and that's it, as they never have time for more testing. If they could generate 10000 broken docs, run the software on each of them overnight, get a set of pages on which it crashed, then fixing each crash bug is 5 minutes of work, no more.