Bug 47967 – Plain Text Representations of XML Source Code without .txt Extension Result in Parsing

• Status: UNCONFIRMED
• Component: Page Loading

[Hugh Guiney]

I saved an XHTML document as a text file so that I could display the source code for users to download. I have Apache HTTPD set up to leave off extensions with MultiViews. So while there is only one resource, xhtml.txt, but it is accessible via /xhtml as well as /xhtml.txt. In either case, the response headers report "Content-Type:text/plain; charset=utf-8". And yet, if I request /xhtml, WebKit renders the document incorrectly, parsing it as if it were actual XHTML. If I request it with .txt on the end of it, it renders the document correctly as plain text. However, if I remove the XML prolog, it is rendered as plain text in both instances. Gecko and Presto both display plain text regardless of whether an XML prolog is present.

[Alexey Proskuryakov]

The draft specification governing content type sniffing is at <http://tools.ietf.org/html/draft-abarth-mime-sniff>. I don't know whether Safari works in accordance with it here.

[Adam Barth]

The draft prevent sniffing of XML from text/plain.  Safari does sniff from text/plain in some cases.

[Alexey Proskuryakov]

The same document may still get sniffed as HTML.

I guess we need to have the original document to find out what exactly happened. Reporter, would it be possible for you to upload it?

[Hugh Guiney]

Created attachment 71322 [details]
XHTML as Plain Text

[Alexey Proskuryakov]

Thanks! Yes, that's sniffed as <application/xhtml+xml>.

Adam, what does the spec say? Should it be text/plain, or text/html?

[Adam Barth]

> Adam, what does the spec say? Should it be text/plain, or text/html?

text/plain.  We should never sniff HTML or XML from text/plain.  The spec says text/plain can only become types that are not scriptable.