Forum OpenACS Q&A: Response to Would copy speed PG installs of reference data?

I understand (and agree with) the reluctance to store the data in multiple forms.  The comma delimited data format can be generated on the fly from the SQL, at least for the case I tested with, just using grep and sed.  The timezone data is perhaps a little more awkward, but surely not impossible.

So there is (I think) no need to store the comma delimited form in CVS or tarballs or RPMs.  Just generate it at install time, before doing the copy or Oracle equivalent.  That way the editable, developer-preferred, format of the data remains the SQL statements as they are now.

It's not the end of the world if we can't do this, but if I automate
things further, so that non-core packages get installed automatically by my RPMs, the delay during the install will grow, and may worry some novice installers enough that they abort it thinking something "must have gone wrong".  I noticed that just a very few packages were responsible for the bulk of the package install time, so I took a quick look, and this seems like a useful way to shorten that delay.