[XML-DEV Mailing List Archive Home] [By Thread] [By Date] [Recent Entries] [Reply To This Message]

Re: collisions (was The subsetting has begun)

• To: xml-dev@l...
• Subject: Re: collisions (was The subsetting has begun)
• From: Joe English <jenglish@f...>
• Date: Tue, 25 Feb 2003 17:13:31 -0800
• In-reply-to: <B885BEDCB3664E4AB1C72F1D85CB29F804B54D50@R...>
• References: <B885BEDCB3664E4AB1C72F1D85CB29F804B54D50@R...>

Dare Obasanjo wrote:

> Exactly what is wrong with the "Project Record in XML" from a document
> design perspective?


The main problem is that all the URLs point to a CGI
redirector on freshmeat.net, instead of the real URI:

    <url_homepage>http://freshmeat.net/redir/dom4j/16508/url_homepage/</url_homepage>

Further, since they're all uniformly constructed
(http://freshmeat.net/redir/$projectName/$projectId/url_$urlType),
the only useful bit of information you can glean from
the entire element is whether or not there exists
a URL of that category for that project.

Aside from that, the document structure looks suspiciously
like a direct serialization of a badly-designed RDBMS schema --
you can just *tell* that the site is backed by a database
containing something like:

    CREATE TABLE project_urls
    (
	project_id char(32) primary key,
	url_homepage char(100) null,
	url_project_page char(100) null,
	url_tgz char(100) null,
	url_changelog char(100) null,
	url_rpm char(100) null,
	url_deb char(100) null,
	url_bz2 char(100) null,
	url_cvs char(100) null,
	url_list char(100) null,
	url_zip char(100) null,
    );

A better design (IMO) would be an XML serialization like:

    ...
    element urls {
    	element url {
	      attribute role string
	    & attribute href string }* }
    ...

where 'role' is one of 'homepage', 'project_page', 'rpm', etc.,
which would be the natural serialization of something like:

    CREATE TABLE project_urls
    (
	primary key (project_id, url_role)
	project_id string references project.project_id,
	url_role string references url_roles.url_role,
	url string
    );


Or going even further, the freshmeat.net database would
be a killer app for RDF.


--Joe English

  jenglish@f...

• References:
  • RE: collisions (was The subsetting has begun)
    • From: "Dare Obasanjo" <dareo@m...>

• Prev by Date: RE: collisions (was The subsetting has begun)
• Next by Date: Re: Parsing efficiency? - why not 'compile'????
• Previous by thread: Re: collisions (was The subsetting has begun)
• Next by thread: RE: collisions (was The subsetting has begun)
• Index(es):
  • Date
  • Thread