On Mar 14, 2012, at 11:02 AM, Barry Leiba wrote:
In fact, I would put it the other way around.
The architecture documentis very useful, almost necessary, for deciding
whether the solution is a good one.
The cache management evaluations are another component of such an
Even understanding the cache management question is made significantly
easier by having a clean architecture description.
At the risk of straining things with an analogy, I'll liken this to
architecture of a building. At the architecture level, we do have to
consider, say, bathrooms, and it's reasonable at that level to talk
about having bathrooms on each floor, and perhaps the extent to which
accessible bathrooms are needed. Deciding whether to have two pairs
or three per floor is something that can be left to the detailed
post-architecture specifications, and things like what sorts of
fixtures to put in them *absolutely* come later.
No one is saying that some discussion of caching issues (etc) won't be
needed in defining and specifying the architecture. We're just saying
that it's not necessary to do the caching document before, as long as
the related discussion that happens at the architecture stage is
preserved for when it's time to work on the caching doc.
Analogies are always risky, but I'll continue. I don't think the proper analogy
is a bathroom. Everyone knows you can build bathrooms that work, having seen
successful examples in the past. You might instead think of a building which
includes a perpetual motion machine [*] as a key element.
[*] Forgive me, press of time prevents me from finding an example which is both
less inflammatory, and still has the property "something whose practicability
is debatable." Feel free to fill in your favorite.