Friday, November 6, 2009

Codes, Designs, and EPI

Today has been a fairly productive day, despite a heck of a lot of interruptions; to the point in fact, that at least 2 hours of work time was lost to it... Thanks ma!

Most of my mental energy was devoted to refining the interface between epi-add and $EPI_WIZARD, and figuring out how best to document it. My original vision for it was using bidirectional communication between finite state machines running in separate processes (the installer and the wizard). Amid the 21,000 interrupts of the day, I've managed to balance out the problem, and have come up with a more interesting solution; one which vastly simplifies the wizard programming interface and grants more freedom to anyone who wants to write an $EPI_WIZARD, which should be easy as pie.

By contrast, most of my code time was spent working on epi-unpack and prototyping ideas for the previous problem. Other then a few changes that I would like to make, epi-unpack is basically done; I'll probably work on epi-verify next, while the others are reviewing the code for epi-unpack. One thing that distinguishes our Encapsulated Package Installer (EPI) system from PC-BSDs PushButton Installers (PBI; formally PcBsdInstallers), is that PBI is a static monolith from the undocumented garage; ours is knit atop a framework of UNIX programs, with standards and manuals to be shipped along with them ;).

I can not lie, UNIX has effected my philosophies on software design—for the better.

Generally, I don't discuses business or classified projects on my Live Journal as a matter of ethics, but since EPI is now public knowledge, I'm free to blog about it's development. The same can't be said of all things SAS or work related lol. Most likely more things will filter through about EPI, so I've created a `tag` for it. Over 3 years and 1500+ entries, and I have never really gotten into Live Journals tagging feature, but have been contemplating it for the last few weeks.

The only way I can ever find my old entries is through Google or sequential search, nether of which are reliable; so utilizing memories and tags would be a darn good idea by now. The problem of categorizing my thoughts, as always remains a problem :=(+).

No comments:

Post a Comment