This bug is a related bug from bug 59411: If you upload a file that is of filetype "unknown", the file is currently set to "application/octet-stream". It could be nice if some kind of byte-scanning were used to determine the mime-type if it's not known. IE does this.

eric I will let you have this for now

RFE cleanup. RFE is already indicated by the Severity field...Sorry for the spam!

Bulk reassigning form bugs to Alex

If we end up using nsIRequest, then we could use existing functionality in ./netwerk/streamconv/converters/nsUnknownDecoder.cpp. Setting dependent on bug 114106 in case we go that route.

any news on that? On Un*x platforms the most files don't have extensions. So Mozilla does always upload application/octet-stream. There are many webinterfaces which don't allow to change the mime-type after upload for the documents.

See ongoing discussion in netscape.public.mozilla.netlib

Short story: this is a question of how we should expose the functionality. Several options: 1) Have a method that takes a buffer and returns a type (essentially expose nsUnknownDecoder::DetermineContentType). Pros: simple. Cons: needs a buffer. 2) Have a method that takes a rewindable stream and returns a type. Pros: also pretty simple. Cons: needs a rewindble stream. 3) Have a method that takes a stream, returns a stream to read from (which therefore has to wrap the other stream), and allows the type to be read off from somewhere (or just returns it). Pros: don't need anything special. cons: rather complicated. In all methods, the nsIFile should get passed in in addition to the stream, so that nsUnknownDecoder can just apply its usual algorithm. biesi, darin, thoughts?