These are observations rather criticisms. I assume that the more thought and discussion I am involved with on this topic the better I will finally understand it.
Let me explain -
The project I work on professionally is a pretty large government funded web site and I have just been working on the 'software refresh' - I think that name reveals plenty about where this is all going. 'Software refresh' really wasn't in the sense that there is a large content driven web site that surface many nodules of discrete functionality through bespoke code. Each nodule is still part of the whole, and the normal way would be to see them as packages that have a set of dependencies that vary according to version.
I think you know where I'm going.
If you substitute perl for the server scripts in ant and shell the situation conceptually may be similar to that of the movabletype codebase?
Perl has CPAN and associated tools, Java also has some tools. But it is very difficult to apply these tools retrospectively.
For instance Byrne keeps old versions of Photo Gallery linked - and this is a different type of decision as to whether the link should be removed, than publishing version and version dependencies for Photo Gallery. It is a different (logical) order of decisi
Image by art_ikon via Flickr
An equivalent example from my professional life would be broken links to public content - how should they be mended? This is in control, and of a refresh remit. But, in terms of different order of decision, another example is where the whole code base of a particular project might be better worked in a different directory structure.
Since we are talking about dozens of discrete projects, you may imagine this is an intimidating proposition. And I have come to the conclusion that no-one quite knows their way around these issues.
It is not quite knowing how to deal with this sort of thing that I am assuming applies to the MT codebase, OS and commercial, as well.
So when I complain that the situation around MT is confusing I, at the same time, imagine that the reason is because the situation is complex.
Now I do have something to add to this though. This is that there is a similarity in approach to creating a large web site where many intrinsic bits of code exist in various scripts and have been placed there over the years in, what is described as, an organic way, to what must happen in a community such as MT, both OS and commercial, over the years.
It seems that the only way out is to either ensure that each atom of change works with every other part of the system or enforce categories of code within which change may take place. At the moment most, Java, Perl or other, do a bit of both. Further, by insisting to the extreme either alternative I don't think there is anyone who would be able to say this would evaporate all the problems I have outlined. However, it maybe that there is such a way. What do you think?
