I think you're probably safe to think it will be the same, but who knows
what might change that?
I think it would be best if you didn't rely on it at all. It really isn't so
hard, just means that youre virtual stream needs to do buffering on it's
own. For example, BizTalk reads the first 100 bytes, but the first line is
250 bytes long. You read 250 bytes into your internal buffer, do the
validation, and return the first 100. Now BizTalk asks for the following 100
bytes, which you can just read from your interal buffer without actually
reading the underlying stream. Now biztalk asks for 100 more bytes, and now
you read the next line which is 150 bytes long, append it to the buffer, do
the validation, and return the last 50 bytes from the first line and the
first fifty from the other.