time management for hackers
Stop all the clocks, cut off the telephone,
Prevent the dog from barking with a juicy bone,
Silence the pianos and with muffled drum
I have been reading time management for sysadmins. I was really hopefull when I brought the book that it would show me some stuff that would help me manage my time better.
Alas, it really isn't much more than the standard time management techniques: getting things done plus a pda. Not that I'm disrespecting Tom Limoncelli, he's spot on the that most sysadmins don't know a lot of this stuff. I wish I'd had this book 15 years ago.
But it's not what I wanted.
What did I want? More detail, less getting things done hand waving.
We kicked off WooMe 2 years ago and mrs.nic and I had our fourth child at roughly the same time. My time management techniques were pushed to the limit. Ok. Past the limit.
I had to do something to keep tabs on the tasks of a 10 strong team of hackers. I started to force more of my stuff through a scripting mill. Trying to script things at least means you are specifying how you think things work. If only to yourself and your shell.
So here's some techy tips, concrete things I've done that have given me more time and more ability to manage the time I have.
Everybody, especially Tom Limoncelli, will tell you that for time management you have to write everything down. You have to record the stuff you've done and the stuff you're gonna do. Let's be clear about this: you're making a bunch of databases. Probably not just one because some of them will be personal and some of them will be shared with other people. But, because you are a hacker it is vital that these databases should be ones that you can hack.
The next obvious thing is that if there are lots of databases you need to try to reduce them to as few databases as possible, those you are left with you have to synchronize where necessary.
The devil is in the where necessary bit.
Out of interest here are my databases:
* the WooMe ticket system, happens to be trac * an emacs org-mode todo and projects file * an emacs org-mode file of my contact database ** I keep all the WooMe tech team in here, mobiles etc... * the WooMe tech wiki where we keep lots of unstructured and structured info * mercurial - all WooMe's code is kept here All these systems are easy to get data in and out of and that's why I chose them. Trac has a great API for getting data in and out. org-mode is just text.
Now the devilish bit: how do I synchronize them? I do things like:
* releases * pulling to be released ticket numbers from trac into org-mode * fiddling with them * exporting them back to trac wiki * making tickets from org-mode todo's * jumping from a source file in emacs to the mercurial version
These are all synchronizations and have been chosen through looking at what I do, not by building a general tool. I find editing ticket lists easier to do in emacs. I find code revisions easier to look at in hgweb than in emacs. I find I keep todo's of tickets for some time polishing ideas.
To me, synchronizing just means not having silos. Make sure you move data around easily.
Time management like any management is overhead. What you do not want is for your time management to be a burden. This is certainly the reason I have resisted adopting formal strategies, I felt the overhead was too great. But this is the key thing that hackers can improve. Hackers automate stuff without thinking, why is it so many of us get stuck with not scripting our time management tools?
I stopped having my own bin dir about 4 years ago, declaring it passe. However, since this year I have started to try writing scripts for everything I do more than a few times a day related to time management. I now have programs as trivial as this:
#!/bin/sh date '+%Y%m%d'
I call this isotoday - it prints the current date in iso format;
#!/bin/sh date '+%A %d %B %Y'
I call this today it just prints the date in human form.
Or even this:
#!/bin/sh date '+%Y%m%d%H%M'
I call this isofull and whenever I make a release of WooMe's code (which is about 2 or 3 times a week) I type:
hg tag release_$(isofull)
It doesn't matter how small the script is. today is ridiculous really. I think the important point here is
@emph don't be shy
These tiny scripts do really save me brain time: typing today instead of the date command probably saves me 10 seconds everytime I do it. So I do it more.
Another example is a simple shell command I have to turn org-mode lists into trac lists. So when I take notes of a meeting I can do it in org-mode and then pretty quickly publish to trac.
The corollary to this is don't spend a lot of time with tools that don't let you use scripts. I don't write a lot of text inside firefox because it isn't a great editing or scripting environment.
Even in firefox there are time saving tools: I have toolbar macros for creating trac tickets and doing trac searches.
making a new ticket - I bind this to "nt"
doing a trac search - I bind this to "ts"
GTD - the bad bits
GTD specifies that you keep 1 copy of your todo list and calendar for every single day.
No way. It always struck me as an absolutely idiotic thing to do. Why would you do that? This is a problem with some of the standard time management advice I think, I've been told these things before and said to myself "I'm not doing that, it's nuts" because it sounded to me like a massive violation of DNRY. In other words I was too stupid to see beyond the implementation. The answer to this one is simple: Hackers have version control.
So I guess I do now keep 365 versions of my todo list. But I do it by checking in my todo list at least at the start of every day, changing the date on the heading (just for diff context) removing all the DONE items and then checking it in again. And yes, I also script this.
If I want to see a previous day I simply use hg log, here's a snapshot:
changeset: 9:89e86a150932 tag: tip user: firstname.lastname@example.org date: Tue Nov 03 04:29:11 2009 +0000 summary: starting the day diff -r d1fc56c13ab7 -r 89e86a150932 todo.org --- a/todo.org Tue Nov 03 04:27:24 2009 +0000 +++ b/todo.org Tue Nov 03 04:29:11 2009 +0000 @@ -78,15 +78,11 @@ * woome ** status -*** 20091102 +*** 20091103 **** TODO switch over mdb to frank's new db ***** consolidated tables ****** will be broken? **** TODO fix up otu's script to pull data off files -**** DONE rate detection into rc -**** DONE add allan's expiry cron to rc -**** DONE make a ticket for seb to generate new images -***** should we drop the existing thumb from the db? **** TODO change the movement of tickets so it doesn't change the owner ***** when a dev pushes a ticket to george for product acceptance ***** when a dev (or george) pushes to kelley for test
The stuff you can't script
Ok. There is stuff I can't script. I can't script concentration. Or clarity. That's up to you.
I am no great concentrator. I have 4 children and a twitter account. I can't concentrate on anything for more than about 5 minutes. But that's normally ok, I make sure that I have 5 minutes with good information. All this stuff about pomadoros and wot not is just macho will power stuff and I don't do it. But every week I get through a whole bunch of releases, keep my team running, deal with problems and write some code. I just do it in 5 minute slots. It's not revolutionary. It's just time slicing.
As I think of more stuff I'll update it here.