WWW Futures

(opened on 26.03.95)

This page contains links on WWW extensions, both proposed and already realized ones. Extensions can take the form of new protocols, new document types (HTML 3.0 and higher), multimedia, internationalization, etc.pp. Sub-sections:


Java

(06.03.95, revised 07.03.95) WebRunner Home Page. WebRunner is supposed to be (the alpha-release has yet to see the light of the 'net) an extensible browser. The extension language, Java, looks and (i suppose) feels like C++, but with some modifications: the underlying object model is secure, there are interfaces, the whole class structure is slighly modified. The idea is this: when WebRunnerencounters a document of unknown type, it can request the Java "drivers" that implement said object's behaviour. Since Java is only a core language, there has to be some API to interface with WebRunner's communication and display facilities. IMHO, these should be the real pièce de résistance, but we'll see (after all, it is only a bunch of graphics libraries that make the difference between PostScript and Forth ;-).

To sell Java as a "scripting language" is IMHO wrong, to me a scripting language is something more at the user (in this case, document) level. It is a "dynamic" language, and seems to lend itself to "extensible programs".

The home page doesn't contain too much material, only a link to three postscript files, one of which is the Java language definition. Sadly, no API definition. You might want to look at comp.lang.misc, where a discussion re: Java/Oak is still going on (started on Feb 23). Quite subdued, since most people are waiting for the alpha release and/or more documentation to emerge.

(26.03.95) WebRunner has been renamed "HotJava". How embarassingly cute.

There is now a documentation directory containing some of the old PS stuff, converted into HTML:

Also, there is now a directory of example applets. Now how typical will these be? Reminds me of the postscript snippets one often sees, that demonstrate how efficiently most anything can be done in PS. In practice, most commercial (and non-commercial) PS-generating tools don't care shit about efficiency and size of their code.

The API User's Guide and Packages Index are probably the most important documents of the whole tree, since they define the resources that are accessible to the applets.

Why do they have to sprinkle those damn "interface index" and "class index" and "method index" and "methods" graphics through the indices? And amidst all these pathetic purpose-less graphics, a page with an ASCII graphics! Oh, the embarrassment. (yes i know, if it were an imagemap, i would rant even more ;-)

Ahh, the Searching the Documentation doesn't actually let you search the documentation (as you naively might think), no, it explains how you can use the search tool of your HotJava browser. What? You are not using a HotJava browser? Sorry, mate.

Anyone care to explain to me what that graphic on the security page is supposed to represent? A canine? The old Enterprise logo? No, it is the JavaLogo(TM, probably)!

Playing games with the term "security": three of the four "security features" are really about ensuring that an applet doesn't crash your browser. The last point "Interface-specific security that prevents applets from doing destructive things." surely depends upon the definition of "destructive".

Again, they explain how much secure Java is by pitting it against C++. Sigh. And they are so proud of their typed bytecode. What has this to do with security, for heavens' sake? Terminal fear "gets"-like bugs, eh?

Finally, "security level four" addresses the real problems. I accept that that the JAVA designers have thought about the technical issues of security (such as file and net access permissions for each new applet/class), but what about the UI problems? After how many "applet wants to access resource `foo'. Allow it?" warning messages will J. Random User disable the security stuff? How many HotJava binaries will be installed with security disabled by operators that don't want to answer countless newbies' questions about the security messages?

Btw, what are the security features for including ``native methods''? (C libraries that pose as Java modules)


(13.04.95)

We now have a local HotJava installation. It is not as bad as i thought. In fact, the basics work well. The URLs for documentation (of the form doc://) work easy and fast; the demo applets are cute (and their sources not too daunting); and the loading asynchronous. Minimal problems (no horizontal scroll bar, base tag is ignored) and annoyances (no keyboard shortcuts predefined, not even the basic arrow-keys for scrolling). Good layout; the four basic functions (forward,backward,home,reload) are at the bottom, and not hidden beyond dozens of hardly-ever-used functions. The upload comments (number of connections currently open, bytes left) are good.

Display and font styles are as good as those of Mosaic, much better than those of NetScape. Some of the NetScape extensions are implemented (center), you can't blame Sun for it. There doesn't seem to be much of HTML 3 support (i traversed the Arena tour, and neither tables nor math extensions did work.

The documentation search tool and the doc itself work well together. Having the doc (especially the "man pages" for the Java library!) locally available surely helps. Sun here avoids the bad mistake of NetScape to assume that everyone can painlessly access the company's documentation on-line.


Beyond Latin-1


If Unicode was the default character set of the WWW, many problems would have been avoided from the very beginning. As it is, it will probably take a dozen years until a sensible character set can gain world-wide acceptance. (If ever -- EBCDIC is still with us, after all.)
Unicode should gather wide-spread acceptance in university and programmer circles once the next Emacs version comes out. It is supposed to include most of Mule's functionality, but using Unicode internally (instead of shift representations, as the current Mule does). Once programmers are used to have greek characters, full symbol fonts, and decent quotes, dashes, spaces and dingbats, they won't tolerate systems that don't support those.

Squabbles over Unicode's reduced Chinese/Japanese subsets are just that: unimportant to most of the rest of the world. Europe, Russia, America, Africa and India couldn't care less. How important are China, Japan, and Korea in the software and publishing industry? (Hardware is an entirely different thing.)


(26.03.95) Mr. Manuel Tomas Carrasco Benitez (of the Commission of the European Communities) has written a page on the multilingual normalization of the Web. He has some interesting applications in mind, especially EC-related multilingual documents (treaties, protocols of multi-national conferences). The idea to reserve a few Unicode code pages for markup purposes does have a certain flair, but is definitely un-html-ish (though it should be possible to express it in SGML).

Beyond/Beside HTML 3.0

Documents types that try to replace HTML, or that address totally different issues.
(29.03.95) Witness this announcement of Netscape embracing PDF.
(31.03.95) News Mosaic versions support the viewing of HDF and netCDF. files. Never heard of either of those before; seems HDF it is a "scientific visualization" formalism, the example formats (it is extensible, but there is of course a core) are raster images, and n-dimensional arrays (fancytalk: "gridded data"). Also, annotations (text labels) on single HDF items or whole files. Now if I could find a single HDF or netCDF file on-line to look at!

Resources:
HDF Information Server (HDF Reference Manual)


Servers, Mirrors, Proxies, Bandwidth


(09.07.95)

The current system of servers, mirrors, and proxies will have to change; it doesn't scale too well. URIs will have to be implemented somehow. Two ideas are Dan Dyer's proposal to facilitate information sharing on the internet (that's the one with the lonely PC in the spinach research office) and Jutta Degener's URL Market.


Etc


This page is part of the Trash Heap.