More fortuitous linkage from J. Udell's weblog - this time indirectly : Dealing with Diversity covers some of the same bases I failed to address properly in a previous incomplete post. Sam Ruby seems to like typing - but in an open permissive perlable kind of way.

I think types are fine - when pragmatic development practices make typing something that simply happens without the conscientious effort of the developer. I think most of the reservations against typing are related to this thinking also: Text-centric computation can make do without typing to a large extent because the ultimate consumer - the reader - can do all the type inferencing machines can only dream of.

Approaches at distributing large scale, open, universal type systems beyond the variable set of C seem to fail a lot - and I think this is mainly because it's just hard work. Strongly typed web-services do not subtract from this work without a new level of service of the systems consuming the data from the services. And even then the general guidelines on data before computation, and open by default data universes are the rules to go by.

This whole thing is of course also related to some of the remarks in this post on explicit versus implicit design, and failure or nonfailure of object design.
While I think the patterns movement and the mode of thought it supports is a usable way out of specification hell, I think explicit generalization in strongly typed languages through complex object interactions are a cost to the developer that is is not entirely clear is justified by functionality or reusability or clarity of form and function.

Once again I think the perlish notion of scripted orchestration of compiled code wins hands down in this respect. In weakly typed languages the orchestration can be quite briefly written, and if there is a well-written interface to strongly typed libraries the confrontation between strong and weak types need not hurt so much.