REQUEST FOR COMMENT: .htmld format

Rohit Khare (khare)
Mon, 6 Feb 1995 18:22:39 -0800


Since the .htmld standard seems to have settled down, I'm forwarding a WebStep
RFC on the topic.

We need some serious input on encoding schemes for upper ISO-Latin-1.

Should we standardize on .htmld or .html for the wrapper extension?

Rohit Khare
---------------------------------------------------

WebStep RFC #1: .htmld Document Format

DESCRIPTION
-----------

RFC #1 specifies the standard content and form of an .htmld document wrapper.
An .htmld document encapsulates all of the resources for a single index.html
document:, including slight variations of format (index-imaged-elements.html)
or name (foobar.htmld/foobar.html). This RFC cites external specifications to
define HTML and OpenStep document management.

RATIONALE
---------

OpenStep relies extensively on the use of "wrappers": .app wrappers, and
application-defined .doc wrappers. While NeXT has registered .htmld as a
standard extension for HTML-based documents, there are several design choices
involved in building an interoperable .htmld

SPECIFICATION
-------------

Foobar.htmld MUST CONTAIN
* index.html

it MAY CONTAIN
* alternate indices of the same html content (index-ascii.html,
index-imaged-elements.html)
* referenced files
* application-specific data (.linkdb, .etDocInfo, other state)

The HTML used MUST CONFORM TO:
* HTML2.0 document structuring conventions
- no <BLINK> tags or obsoleted tags (<XMP>)
- valid <HEAD> elements
- HTML2.0 entities
* HTML3.0 entity maps for upper characters and foreign languages
- What about Symbol font?
- What about NeXTSTEP encoding fallbacks?

The Component Data, if provided SHOULD CONFORM TO:
* GIF or XBM inlined images
* 8-bit mu-law voice audio
* Future applicable WebStep standards for document description

[The following section may be controversial and may be dropped]
STYLE GUIDELINES:
* translated media types (downsampled audio, eps->gif) should be
accompanied by the original data (clicking on gif returns .eps)
* All html resources should be index*.html and linked to each other
* All internal links should use relative adressing

CONFORMANCE TESTING
-------------------

For .htmld authoring tools:
* Output hmtl must conform to strict HTML2 DTD checking
* Documents should be movable across filesystems and machines

For browsers/clients:
* Must be able to "open" .htmld by tacking on "index.html", not by
displaying the .htmld as a directory listing.

EXAMPLE IMPLEMENTATIONS
-----------------------
Pages1.7 app is conformant
eText .95 will be conformant (change extension from .htmd to .htmld)

Most web servers (CERN, NCSA) are conformant (automatically tack on
index.html)

It is not known if OmniWeb or SpiderWoman is conformant.

NeXT's .htmld appears to use a conflicting naming scheme

StepWise .htmld does not support the "one-document" guideline, but is
strictly conformant.