A strategy for more frequent posts

29 Jan 2020 Tags: Distributed systems

I have many incomplete threads on distributed systems. How do I balance my desire for complete and correct presentations with the need to post more frequently?

A year ago, I began two threads on distributed systems. The first focused on using the TLC model checker (and by extension other model checkers) to explore failures of distributed algorithms. Before I could get far with that, I set it aside to begin a second on representing Howard and Mortier’s work on generalized distributed consensus in PlusCal and analyzing it using TLC.

I have a related thread that has taken months to begin, guiding students through the code for TenCent’s implementation of Paxos.

This spring, teaching distributed systems to professional master’s students led me to conceive a fourth thread refocusing attention away from CAP and back to the Fischer-Lynch-Paterson impossibility result.

That’s a lot of dangling thread. Each would require many hours and many posts. Instead of pursuing any one, I turned instead to writing shorter, less dense and more tentative posts about topics of immediate interest, particularly teaching, as that is my primary activity for the next few months. I wanted to regain the freedom to think aloud, which I had to restrain while writing posts that aimed for a more complete statement. However many flaws remain in the published versions of those posts, I removed many more by repeated passes before publishing them. The result satisfied me but the process exhausted me and was partly responsible for my five-month absence from posting.

In coming posts, I aim to balance between these two styles. In particular, I will return to those earlier threads but accept some less precise expressions, in interest of producing drafts to develop the ideas. Once I have collected enough draft sections, I may return to do a more polished version, detailed and precise, a more final form of those ideas.

Or maybe not, if I judge the first drafts to be good enough. My personally urgent task is to break my current deadlock by relaxing the requirement that posts be as close to definitive as I can make them and instead accepting imperfections as the price of timeliness.