Dear Paul, Hans, Matthew, and Kingsley, (01)
The advice about not throwing good money after bad is sound. But the
issue of what to do about legacy systems is very different. Those are
*not* bad systems. On the contrary, the investment in creating them
has been so successful that entire businesses or even the world economy
depends on them. But over time, all systems require changes to adapt
them to new requirements. The major question is how to make those
changes in a smooth transition instead of an abrupt upheaval. (02)
PT
> Still it seems strange to me, but it does explain spending decisions
> that might appear irrational to an unenlightened observer. (03)
There is a large amount of literature about the costs and tradeoffs
of various strategies. The most successful transitions continue to
operate both the old system and the new one for long periods of time.
This is true for hardware and software in every branch of technology. (04)
PT
> I will rebrand and move the thread about app-specific OIDs vs. URIs over
> to ontology-summit, as I believe that is where John's motivating
> questions belonged. (05)
The discussions for the ontology summit are focused on what topics
and issues should be discussed in the April meeting. The issues
about various kinds of IDs won't be resolved there. (06)
HP
>> Could you elaborate a bit on what you mean by "this will already be
>> tied up in system identifiers for free"? (07)
MW
> The data will be in existing systems using relational databases
> (almost certainly) and will have identifiers already as a result. (08)
That is true for a huge amount of critical data in the world. The
technology of RDBs has been tested and proven over the past 40 years.
It's not perfect, but for many purposes it works very well. (09)
Any proposed technology that does not support smooth interoperability
with mainstream IT will be ignored by mainstream IT. Five examples: (010)
1. The overwhelming number of commercial web sites on the WWW
incorporated relational DBs. They succeeded brilliantly on
both the large scale (the kinds of applications Matthew was
discussing) and the small scale (the LAMP technologies of
Linux, Apache, MySQL, and Perl, Python, or PHP). (011)
2. The Semantic Webbers attempted to make a clean break with the
past. They introduced radically new notations and techniques
that had never been tested in practice, and made no provision
for interoperability with the technologies that made the WWW
successful. Compared to the following three technologies,
the SW tools never broke out of "niche" status. (012)
3. PHP was developed by one programmer, Rasmus Lerdorf, for his
Personal Home Page. He started in 1994, the same year when
Tim B-L proposed the Semantic Web at the first international
W3C conference. In 1995, he released it to the world, and its
usage took off like a rocket. In a few years, PHP overtook
Perl and Python as the most widely used P in LAMP, and it is
far and away more successful than any tool from the SW. (013)
4. The most significant technology that made the WWW successful
was the Mozilla browser, which was the foundation for Netscape
and Internet Explorer. It pioneered the use of <script> tags
for incorporating languages (such as JavaScript). And JavaScript
adopted many of its data conventions from Java, which is another
popular language for the WWW. JSON (JavaScript Object Notation)
was developed by Netscape, the same company that developed RDF.
Today, JSON has become the de facto alternative to RDF because
it has a smooth integration with every major programming language. (014)
5. JavaScript was widely used, but with competing variations.
ECMA harmonized the variations in ECMAScript, which became the
foundation for AJAX (Apache, JavaScript, and XML). After Google
demonstrated the power of AJAX with Maps and Gmail, it very
rapidly become a huge success. Although XML is used in AJAX,
JSON is even more popular. RDF is not. (015)
Anybody starting a new business is under no constraint to "throw
good money after bad". They just want something that works.
LAMP works. PHP works. JSON works. AJAX works. But the SW
tools are still niche technology. (016)
R.V. Guha and Tim Bray, the two designers of RDF, worked at Netscape
at the time. They could have adopted JSON as the foundation and
used the <script> tag to include it in web pages. But the W3C
edicted the use of the XML notation. That was a huge blunder.
It was a "proactive standard" that had not been tested in practice,
and even the two designers admitted that it was a mistake. (017)
KI
> Are you saying that using triples or quads are bad period, or that
> they are bad when applied in "rip and replace" mode? I think, but
> do confirm, that you mean the latter. (018)
Yes. There are useful applications for triples, quads, monads,
dyads, etc. But there is no such thing as a one-size-fits all
data structure or knowledge representation. The worst possible
mistake is to force a single representation -- especially when
there is no period of testing, evaluating, or even considering
alternatives. (019)
KI
> Basically, rather than having to choose between Intensional (relational
> property graphs) vs Extensional (relational tables) solutions, we can
> look to a newer generation of RDBMS solutions that offer exploitation
> of both RDBMS aspects, subject to application requirements.
>
> http://queue.acm.org/detail.cfm?id=1961297 -- this article explains
> the point very well. (020)
That's a good article. I would also add that the acronym NOSQL has
been interpreted as "Not Only SQL". Even in the late 1970s, many of
the RDB developers -- including the chief guru, Ted Codd -- said that
SQL required some major revisions and extensions, including a proper
system of types. But new versions must interoperate with the old. (021)
I would also add that the idea of universal identifiers has been
recognized as important for about 500 years -- Descartes, Newton,
and Leibniz were among the many who debated the issues. URIs are
a useful contribution to that debate, but they are certainly not
the last word on the subject. (022)
John (023)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (024)
|