Nice job. Couple of points:
<!ELEMENT other (#PCDATA)>
<!ATTLIST other title CDATA>
What is that for?
The email client I'm using has a line called Subject.
That is specified in:
<!ELEMENT headers (from, to+, cc*, date, subject?, other*)>
but not defined as an element.
Jack
From: Eugene Eric Kim <eekim@eekim.com>
> I've posted a very preliminary DTD for e-mail at:
>
> http://www.eekim.com/ohs/devel/email.dtd
>
> Comments and suggestions are definitely encouraged. I'm going to try to
> whip up some code for converting RFC822 e-mail into this XML format before
> our meeting today.
>
> There are several issues -- some stylistic, some content-related -- that
> arose from developing this:
>
> 1. E-mail is more of a container than a document format. The OHS should
> be able to support any document type sent or to be sent via e-mail. This
> raises the question: How does one specify mixed XML within a document?
> I believe this is a question that was raised earlier on this list.
>
> 2. I developed a very simple schema for the body text of RFC822 e-mail
> that should be useful for transcoding and views, but not much
> else. There's basically one tag -- <p> -- which sets XML's xml:space
> attribute to "preserve" by default -- the equivalent to HTML's <pre>
> tag.
>
> I defined this one tag rather than a multitude of structural tags because,
> after analyzing a bunch of e-mails, I realized that any transcoding
> algorithm that tries to be too smart -- no matter how sophisticated --
> will never be 100 percent right. When it's wrong, you can lose
> or misinterpret structural information, a bad thing. On the other hand,
> raw ASCII text contains a bunch of structural information usually
> specified by whitespace.
>
> I feel that it is better to add some additional structure -- mainly for
> addressing and views -- and to retain the textual structure already
> there. This is something we can discuss in more detail.
>
> 3. Automatically identifying and transcoding citations is going to be
> interesting. I think it's okay not to be 100 percent perfect for this
> type of thing, but regardless, it will still require some cleverness.
>
> I created two simple links: <citation href> and <a href>. Both should
> probably be extended links, but for the purpose of just getting something
> usable out there, I made them simple for now.
>
> 4. One style issue that arose was whether or not to wrap metadata
> information in <header> tags. I opted to do it primarily for readability,
> but this is by no means set in stone. I'd be curious to hear people's
> opinions on this matter.
>
> 5. Looking over the DTD, I realized I left out tags for categorizing
> e-mail. I'll add those in the next rev.
>
> -Eugene
This archive was generated by hypermail 2.0.0 : Tue Aug 21 2001 - 17:57:48 PDT