Web design choices
For MIT 6.001 only...a historical overview of the technology with an emphasis
on design choices and lessons for how we as a race can achieve things.
Myths which existed and endure; Reality, perceptions and their affect on the
growth of the system.
Check out:
WWW Design Choices
We look at some of the historical design choices in W3,
the technical and political effect they had,
and wonder what would have happened otherwise.
Requirements
A global shared information space
- Buy-in from HEP, Internet communities
- Cross-platform interoperability
- Ease of entry of data
- No constraint on topology of information
- Minimal constraint on users
- Evolbability
Web architecure
Hypertext model
- Hypertext rather than menus & objects
- Bidirectional links
- No link consistency
- Links transmitted as part of object
URLs
http://www.w3.org/pub/WWW#Specifications
- Universal
- Opaque
- Not name or address
- Fold in old and new spaces
The Name myth
Myths:
- URLs change
- Names don't
- We need a system to translate from names to URLs
The properties of a name are social as well as technical
HTTP: RFC822 pros and cons
- + Human readable
- + Accepted in Internet community
- - No cleanly defined syntax
- +- Unrecognised tags ignored
HTTP: RPC pros and cons
- + Tools, well-defined syntax, efficiency
- - Little asunchrony or pipelining
- + Easy standardization
- - Big buy-in cost: compilers, runtime.
- - Competitive arena.
TCP pros and cons
- + ubiquitous (cf. DECnet, ISO TP4)
- - slow start (cf. UDP)
- + API on most platforms (cf. UDP)
SGML pros and cons
- Empty tags problem
- Line ends and spaces
- Profiling
- Lack of modularity
- Lack of BNF grammar
- The only standard in the hypertext community
- Great strength in the documentation community
Preconceptions
- Hypertext is confusing
- HTML is too complex
- SGML parsing is a batch job
The winding road to adoption
- WWW and HEP via phone books and VM
- The gopher challenge
- The URL debacle
- HTML and SGML
- IMG and Mosaic
- NaviPress etc
The slow bang
"The web aboutt he Web" info.cern.ch http server hits:
- Internet reduced time constant for change
- Remarkably steady exponential
- Effect: Freezing of the technology
- Differing perceptions at different levels
Postconceptions
- URLs are filenames
- Links must be stored in HTML files
- HTML is an alternative to SGML
- Clients are browsers
- You have learn HTML
- HTML is too simple
The present
- Fragmentation tension
- Time constant for change has shunk
- The cool:
The future
- Vortexes:
- Standard API for agents
- IPR and metadata mechanisms
- The revolution:
- OOP with Global inheritance
- Global proof: the "Oh, yeah?" button
- Serious agents.
and the time constant shrinks again...
The W3 Consortium
- A meeting of 100+ companies
-