The HTTP standard says:
If this header [Content-Disposition: attachment] is used in a response with the application/octet-stream content-type, the implied suggestion is that the user agent should not display the response, but directly enter a `save response as...' dialog.
I read that as
Content-Type: application/octet-stream
Content-Disposition: attachment
But I would have thought that Content-Type
would be application/pdf
, image/png
, etc.
Should I have Content-Type: application/octet-stream
if I want browsers to download the file?
No.
The content-type should be whatever it is known to be, if you know it. application/octet-stream
is defined as "arbitrary binary data" in RFC 2046, and there's a definite overlap here of it being appropriate for entities whose sole intended purpose is to be saved to disk, and from that point on be outside of anything "webby". Or to look at it from another direction; the only thing one can safely do with application/octet-stream is to save it to file and hope someone else knows what it's for.
You can combine the use of Content-Disposition
with other content-types, such as image/png
or even text/html
to indicate you want saving rather than display. It used to be the case that some browsers would ignore it in the case of text/html
but I think this was some long time ago at this point (and I'm going to bed soon so I'm not going to start testing a whole bunch of browsers right now; maybe later).
RFC 2616 also mentions the possibility of extension tokens, and these days most browsers recognise inline
to mean you do want the entity displayed if possible (that is, if it's a type the browser knows how to display, otherwise it's got no choice in the matter). This is of course the default behaviour anyway, but it means that you can include the filename
part of the header, which browsers will use (perhaps with some adjustment so file-extensions match local system norms for the content-type in question, perhaps not) as the suggestion if the user tries to save.
Hence:
Content-Type: application/octet-stream
Content-Disposition: attachment; filename="picture.png"
Means "I don't know what the hell this is. Please save it as a file, preferably named picture.png".
Content-Type: image/png
Content-Disposition: attachment; filename="picture.png"
Means "This is a PNG image. Please save it as a file, preferably named picture.png".
Content-Type: image/png
Content-Disposition: inline; filename="picture.png"
Means "This is a PNG image. Please display it unless you don't know how to display PNG images. Otherwise, or if the user chooses to save it, we recommend the name picture.png for the file you save it as".
Of those browsers that recognise inline
some would always use it, while others would use it if the user had selected "save link as" but not if they'd selected "save" while viewing (or at least IE used to be like that, it may have changed some years ago).
Success story sharing
attachment
could be considered "best not to display this yourself" whileinline
as "best to display this yourself if you can". Either way, most browsers will use the filename value as the suggested name of the file, but users can always override that.application/octet-stream
Content-Type
for downloading files. You don't. They also recommend to setattachment
toContent-Disposition
. You don't. Your explanation makes me feel clear and comfortable. Theirs don't.text/csv
but it could not be opened. It turned out, HTTP client is doing some additional re-encoding and replaces BOM with garbage. The only way to prevent it from doing so was to specifybinary/octet-stream
. I did not tryapplication/octet-stream
yet becausebinary
seemed to be more explicit about what I wanted to achieve.octet-stream
so that a download window can pop up