it occured to me that we could probably eliminate the Accept-Charset and Accept-Language headers from our image requests. i wouldn't expect servers to care about locale specific information when serving images. they would probably use different URLs if images needed to be localized (i.e., the HTML document would be localized to reference different images). of course i'm not sure how to prove my theory, but eliminating these headers would cut back on a significant number of bytes per request: Accept-Language: en-us,en;q=0.5\r\n Accept-Charset: ISO-8859-1,utf-8;q=0.7,*;q=0.7\r\n that's 81 bytes by my count that most servers couldn't care less about. thoughts?
Severity: normal → trivial
Status: NEW → ASSIGNED
Priority: -- → P5
Target Milestone: --- → mozilla1.3alpha
What about people doing text-as-image (or svg ;)? Or 404s with a localised ErrorDocument?
>What about people doing text-as-image (or svg ;)? imagelib currently barfs if it does not recognize the data as some type of image format. how does SVG work? do we have a imagelib decoder for SVG or something? >Or 404s with a localised ErrorDocument? i don't follow... don't 404's usually result in a HTML document? or are you saying that websites serve up a 404 image that could be displayed inside a <img> tag? i suppose they could. have you actually seen a website do this before? BTW, i wouldn't be changing anything for toplevel image documents... i'm only talking about inline images (<img> tags).
>What about people doing text-as-image (or svg ;)? I think by "text-as-image" Bradley meant a representation of some text in the form of an image (eg scanned-in document). If the original document is available in multiple languages, for example, Accept-Language may make sense as a result. svg is not handled by imagelib at the moment, and really never should be (it should just be embedded directly in the page).
We don't have an imagelib for svg yet - note the ;) bz: The svg spec requries it to be usable via <img> If someone has an <img> link to a valid document with ahs a 200 return code (by mistake) and then the user loads that document in a window, won't you get the cached version, w-o accept lanugage? You could play games with the Vary header, to detect that, I suppose, although this is an edge case.
hmm.. so maybe this bug should be closed then. it seems wasteful to send these bytes each-and-every when they are ignored 99% of the time. oh well :( if anyone has a convincing counter argument, please feel free to reopen.
Status: ASSIGNED → RESOLVED
Last Resolved: 16 years ago
Resolution: --- → WONTFIX
I was playing Devil's advocate ;) I think we can drop Accept-Charset, though, since theres no charset for a picture. Thats assuming that the caching I mentioned isn't an issue, though.
hmm... but SVG can be used to render text, right... so wouldn't the charset be relevant to SVG content?
Yeah, XML (including SVG) charsets are best delivered over HTTP. They can be delivered in the XML declaration, but that's really just a refined version of the <meta http-equiv> kludge.
Verified wontfix per comments.
Status: RESOLVED → VERIFIED
QA Contact: httpqa → junruh
You need to log in before you can comment on or make changes to this bug.