[Syonyk]

That seems to be how one of the exploits from a year or two ago worked.

https://googleprojectzero.blogspot.com/2021/12/a-deep-dive-i...

It exploited an archaic Xerox format parser to make its own virtual machine, and then went out from there.

So I'd agree, throwing anything on a webpage (or incoming message) into the "Can you parse this weird thing?" pipeline is a bad idea!

[astrange]

JBIG2 is a mandatory part of PDF, not its own weird image format. (Though I think it's also allowed in TIFF files and those might count as weird.)