Lists Home |
Date Index |
You might not get this. The answer is the next
paragraph. The rest is rant. len is cryptic
most often because he is arguing with himself
while replying to someone else. It's a fatal flaw.
I retitled this reply so my rant doesn't dilute
your technical thread.
I'm saying that any change to a namespace declaring
document (schema, XSD, natural english), is a change
to the namespace. You can version number it or change
the namespace declaration. I prefer the namespace
change because it requires no special code to detect it.
The reason to change is to track a variation even if
only within a dyadic relationship. On the other hand,
code that runs escapes such and that becomes a problem
for new parties. It should be easy and require little
to no custom code to detect a message variant automatically.
Use the least powerful means you can.
I know business moves fast. So does kudzu. When
it was used to keep soil from eroding, it worked
for the purpose for which it was applied. When
it grew across the soil meant for other crops,
it became an opportunistic weed.
One thing that namespace declarations enable: if
there is some threshhold number of them, the system
is signalling that it is time to revisit and
shrink the number of variants, or to bifurcate because
these are two different semantic ecosystems even if
they have hot ecotones.
Farming is not about getting maximum yield from land
to enable the most mouths to feed. That leads to
monocultures and monocultures inevitably destroy the soil.
Farming is about growing a mix of crops from season to
season that feed as many as can be fed while preserving
the lifebearing properties of the soil so as many as
can be fed can be fed for as long as they can be fed.
Standards can become monocultures. That is why the
hard work on XML was to get extensibility into the
design whereas SGML emphasized validation. That shift
did not devalue validation; it increased its value
by moving the system away from being a monoculture-accelerant
to being a controlled inhibitor.
Farming is, like any business, what you make of it. There
is no affective reality where we aren't actively setting
the conditions. The universe is full of signal but we
spend millions of dollars looking for just a few. Why?
Intelligence is not completely subordinate to the forces
of the universe and that is why it stands out. That is
either a deep mystery or a transparent truth.
From: Fraser Goffin [mailto:firstname.lastname@example.org]
Sorry Len, this is a bit criptic for me (except the last part :-). I can't
tell if you're agreeing with me or Ken or neither of us ?.
Later on in this note I am also trying to assert that there is a difference
between a minor (non breaking) and a major (explicitly breaking) change and
that a vocabulary that provides no extensibility mechanism (for non schema
authors as well as schema owners) won't be flexible and fast moving enough
for business agility (I hate that term - but its late so it will suffice).
If it comes to a choice, the sacrosanct that Ken is talking about with UBL
will be sacrificed in a heart-beat. I'm not saying thats right, I just
saying, thats reality. So IMO we need to work extra hard to avoid that
condition appearing and if that means compromising (slightly) and not being
a slave to a difficult spec and a slow moving change process (show me a
standards effort that isn't !), then so be it. I have tried (v.hard) at
advocating the purview line and, as I'm sure you are well aware it is a
tough sell in the face of [short term] business delivery demands.
Do you have a view on this (silly question :-)
>From: "Bullard, Claude L (Len)" <email@example.com>
>To: 'Fraser Goffin' <firstname.lastname@example.org>, gkholman@CraneSoftwrights.com,
>Subject: RE: [xml-dev] Versioning of Enums
>Date: Mon, 13 Feb 2006 16:22:36 -0600
>A version-controlled namespace tactic that disallows minor
>changes without version number changes explicitly disallows
>the use of the standard channel to create a tipping point
>advantage covertly. This favors the strategy of openness
>and transparency of transactions and these in turn,
>If in search of signal, I understand markup markets.
>Cheating in markup is like pissing in the party punch.
>Other bowls are provided for that: namespace URNs.
>If the pun escapes you, say it out loud.
>len (ogee, mr grodzins!)
>From: Fraser Goffin [mailto:email@example.com]
>I admit to being in the camp that says the addition of a new value to an
>enum does NOT warrant a new schema namespace (subject to the usual caveats
>about semantic coherence) ?