3 ms·

The thread is fascinating. What I love about it is how the discussion quickly became a variety of proposals for other generalized--and more complicated-- const

by wglass 8y ago

The thread is fascinating. What I love about it is how the discussion quickly became a variety of proposals for other generalized--and more complicated-- constructs. Lots of discussion on making a generic Mime-based embedded object for example. The worst was "<!ENTITY ICON6 SYSTEM "http://blah..">&ICON6;" http://blah..">&ICON6;" Then Marc just ignored the comments, put the <img> tag into Mosaic and it became canon. Would HTML have been so easy for the masses if it was horribly generalized? I'd argue the simplicity of tags like <b>, <img>, <a> made it so popular.

goto11 8y ago
Some suggestions are horrible, but some of the suggestions are pretty good and forward thinking. Allowing img elements to have fallback content would have have the introduction of new image formats (jpg, png, svg) much more painless. We later got the alt-attribute for fallback text, but just using the content for fallback would be both simpler and more powerful. It is probably a case of worse-is-better. They chose the simplest possible solution and shipped. Introducing the img tag is what made Mosaic dominant. Later Netscape went to far with worse-is-better. Remember they introduced the <SPACER> tag to allow designers to control whitespace, at a time when Internet Explorer already had shipped with support for CSS padding and margins. Worse-is-better only works when you ship features before the competition, it does not mean worse features will win out over better features.
- krapp 8y ago
  A generic <include> tag for text/hypertext could have been interesting, particularly if it could be used with link anchors. Although the potential for abuse when including remote content would obvious.
tannhaeuser 8y ago
As an astute reader, you rightfully criticize the use of <!DOCTYPE HTML [ ... <!ENTITY ICON6 SYSTEM "blah.."> ... ]> <HTML> ... &ICON6 </HTML> which doesn't make sense as it declares and then references a parsed entity containing HTML. Of course, what should've been used instead is an SGML data entity: <!NOTATION GIF SYSTEM> <!ENTITY ICON6 SYSTEM "blah.." NDATA GIF> Interestingly, the discussion also evolves around the anchor tag vs allowing href attributes everywhere to make any element work like a hyperlink, as MathML did. The decision to go with anchor tags complicates HTML to this day. HTML5 uses the (unproven) concept of "transparent content" to specify that the content model of <a> is inherited from the context where <a> is inserted [1]. Another interesting tidbit here is that <img> is modelled as SGML EMPTY element (HTML5 "void" element), a decision that would later complicate integration of XML/XHTML which doesn't support elements without end-element tags. [1]: http://sgmljs.net/docs/html5.html#transparent-content http://sgmljs.net/docs/html5.html#transparent-content

by wglass 8y ago

goto11 8y ago

Some suggestions are horrible, but some of the suggestions are pretty good and forward thinking. Allowing img elements to have fallback content would have have the introduction of new image formats (jpg, png, svg) much more painless. We later got the alt-attribute for fallback text, but just using the content for fallback would be both simpler and more powerful. It is probably a case of worse-is-better. They chose the simplest possible solution and shipped. Introducing the img tag is what made Mosaic dominant. Later Netscape went to far with worse-is-better. Remember they introduced the <SPACER> tag to allow designers to control whitespace, at a time when Internet Explorer already had shipped with support for CSS padding and margins. Worse-is-better only works when you ship features before the competition, it does not mean worse features will win out over better features.

krapp 8y ago
A generic <include> tag for text/hypertext could have been interesting, particularly if it could be used with link anchors. Although the potential for abuse when including remote content would obvious.

tannhaeuser 8y ago

As an astute reader, you rightfully criticize the use of <!DOCTYPE HTML [ ... <!ENTITY ICON6 SYSTEM "blah.."> ... ]> <HTML> ... &ICON6 </HTML> which doesn't make sense as it declares and then references a parsed entity containing HTML. Of course, what should've been used instead is an SGML data entity: <!NOTATION GIF SYSTEM> <!ENTITY ICON6 SYSTEM "blah.." NDATA GIF> Interestingly, the discussion also evolves around the anchor tag vs allowing href attributes everywhere to make any element work like a hyperlink, as MathML did. The decision to go with anchor tags complicates HTML to this day. HTML5 uses the (unproven) concept of "transparent content" to specify that the content model of <a> is inherited from the context where <a> is inserted [1]. Another interesting tidbit here is that <img> is modelled as SGML EMPTY element (HTML5 "void" element), a decision that would later complicate integration of XML/XHTML which doesn't support elements without end-element tags. [1]: http://sgmljs.net/docs/html5.html#transparent-content http://sgmljs.net/docs/html5.html#transparent-content