FORMAT-SPEC: .htmld

To: webstep@mail.xent.caltech.edu (webstep@mail.xent.caltech.edu))
Thu, 26 Jan 95 12:40:48 PST


Proposal 0: Define .htmld
====================================
Dan mentions that .htmld has been registered as an official document type. The question before us is:

name.htmld/index.html vs. name.htmld/name.html

*) Pages, eText, and default cern-httpd is happiest with index.html
*) Scott's manual StepWise management uses index.html
*) Nobody is considering TXT.html; FWIW, Mail.app uses index.rtf
*) Dan supports name.htmld/name.htmld for downloading purposes.

How can we accomodate either/both? First, what are the circumstances for downloading? D/L the html, and all you can have is text; graphics, charts, &c are "left behind"; indeed, Pages, eText, and StepWise docs are not self-sufficient in that way.

Can a compromise be met by using a symbolic link within Dan's server? I think if Dan wrote index, linked to name, and proceeded to use the name link in other uses, it would make for a formally interoperable spec. Feel free to dissent; it may be overkill. Of course, if there are no other "essential" files within name.htmld/name.html, why not leave it at name.html, and leave the htmld type for *self-contained* html resources?

2) What about Scott? The implicit rule of defining htmld is "one html per document", like rtfd. However, what about versioning, like a possible index and index-with-graphics approach? Should we define the standard to be "the same html, in possibly several files"?

Of course, we should address why we should standardize on this now. I think that in the future, we will want tools to browse and edit native htmld, so it would be useful to define a standard name for the internal file.

Finally, if I haven't ranted enough today, let me mention what my code looks like from the other side of the looking glass. I have TeXD documents I produce with laTeX code. I expect that people will have to edit and modify such .texd documents by hand. Thus, I use name.texd/title.tex, where title != name, and it's a mess, since I can only find the tex later on by scanning for .tex files in the directory. Leave the user-visible name on the "outside", and keep machine-readable documents consistent "inside". The normal code shows up in how eText "opens" .htmd (which it owns, right?): eText tacks on an index.html and bounces it back to the Workspace.

Action Item
----------------
In the light of this discussion I propose moving eText to name.htmld/index.html, and scrapping eText's private htmd designation. Sound cool? There seems to be no need for both htmd and htmld.

Sorry if I'm boring the list.... more to come. Another part of htmld might be a requisite .docInfo that indicates how to access the "real" native-format document this htmld version is shadowing.

Rohit