Issue with description of WebM sniffing

[velezbeltran]

Hey!
I am currently implementing a version of the specification and I noticed something odd in one of the algorithms. In section 6.2.2 I think that step 6.5 I think that it should be greater than instead of less than.
Imagine for example a very large resource. You only require the beginning of the resource to determine the mime-type. As it is implemented right now, this is saying that if you are looking at the beginning you should break. Which is not correct. You should only break if you are too close to the end.

[Gallaecio]

I’m not sure that’s wrong.

The current algorithm seems to tell you to find number size within the first 38 bytes, and then if number size takes you to 4 bytes before the end, and the end is "webm", return true.

Without knowing the WebM standard, the algorithm does not sound illogical to me.