- From: Robert Burns <rob@robburns.com>
- Date: Mon, 20 Aug 2007 11:28:02 -0500
- To: James Graham <jg307@cam.ac.uk>
- Cc: Julian Reschke <julian.reschke@gmx.de>, Dan Connolly <connolly@w3.org>, "public-html@w3.org WG" <public-html@w3.org>
Hi James, On Aug 20, 2007, at 10:55 AM, James Graham wrote: > > Julian Reschke wrote: >> It would be really nice if there'd be a simple way *for us* to get >> a feeling how big of a problem this is in practice. So I'd really >> like to have a browser that allows me to opt-out of sniffing, or >> that minimally informs me about these kinds of problems. > > It seems like this wouldn't be too hard to gauge by implementing > the algorithm as specced as spidering the web for documents labeled > as text/plain and looking for the fraction that are not sniffed as > text/plain by the algorithm. > > In general I would not expect any mainstream browser to expose an > option to turn off content sniffing in the UI since it would not be > understood by the vast majority of users. I understand what you're saying. I think many browsers would be reluctant to do that. However, I could see allowing users to change the MIME type treatment of a resource after its loaded. This would be especially useful for those cases where content-type header was meant by the author to be authoritative and sniffing gets in the way. A browser could provide a contextual menu on embedded content and a popup menu in a table view of resources to expose a selection between the authoritative type and the sniffed type. A third selection could enable an advanced view to select any IMIME type the UA could handle. To me this would not be any more intimidating or unusual than having a list of text encodings that most users have no idea what to do with. Right now every browser, even Safari (which tries the hardest to keep it simple) has this long intimidating menu to handled charset encoding. Again, most users have no idea what to do with that. You have to have some sense of what an encoding is to select an encoding. How many English speakers know they should select something with Latin in the name? Or even worse, in terms of recognition, UTF? Having a menu selection of two items strikes me as rather simple in comparison. This could even allow a mistakenly interpreted CSS file to be applied to the document. Or enable a script that was treated as a text file. This is something we could recommend to UA implementors with a SHOULD norm: in other words provide the UI needed to make this selection on all resources. Secondly, related to this is the issue Sam Ruby has raised regarding viewing XML (and this could just as easily apply to HTML) as a tree or with styles applied. I don't think any HTML recommendation ever included norms to show the source text of an HTML file, but it became the standard practice. I think it would be a good idea for us to include a norm that UAs SHOULD allow users to view the content as an document/element tree as browsers now do with unknown XMLs. This seems like a more valuable feature than viewing the raw source. The other option already mentioned, is to add either a new content header or a new flag to the existing content header to insist this is a more authoritative header. Then we'll try not to mess that one up. Add me to the list of support for that idea. Though I agree that should be done through the HTTP WG. Take care, Rob
Received on Monday, 20 August 2007 16:28:17 UTC