trying to stay awake for the final hr of pre-con session. damn you jetlag... #pdc #pdc08 #pdc2008
#pdc msft web-devs beware. new pipeline to know about (iis7) update those pre-canned interview questions
#pdc google "visual round trip analyzer" very nice front end for netmon - much more detailed than firebug
#pdc showing a firebug-like round-trip viz tool
#pdc "web application performance is primarily constrained by network latency" (insert snarky captain obvious remark here)
#pdc "last time i heard, we have 500 million hotmail mailboxes"
#pdc uh oh new pre-con web guru speaker just said "html protocol" twice in a row. hoping it was just a brain cramp...
#pdc just announced that pre-con sessions are *not* recorded! (sl2 session was tho, at least when i was there)
#pdc will definately be checking out tpl (task parallel library). sounds like java.util.concurrent in java5 (sane parallel api)
#pdc vadump: drill down a bit more on your working set (google "vadump")
#pdc msft considering a "run this at compile time" attribute. think regex and other codegen. interesting
#pdc perf pre-con going downhill fast with speaker #2. thinking about bailing to the sl2 session...
btw htc (and/or goog), is it really a "mobile" phone if one must spend half of the day tethered to a charger?
#pdc yikes. generic activator.createinstance is 8x slower than non-generic version (they are fixing it)
#pdc disable cpu throttling (ctrl panel power options) when analylzing perf
#pdc pretty sure current speaker just said "shittin bits"
#pdc look at the api differences between non-generic and generic collections - chance to "redo" for perf
#pdc "beware of using collections as your flagship storage - they are not the most frugal"
#pdc 10m statically reachable allocations (largest in framework) is... datagrid.onkeyleft/right
#pdc rico's frank answer to biggest perf problem in the bcl: the config system (right on)
#pdc "in order for code to be reusable, it must first be *usable*"
#pdc heh "object-oriented" vs "work-oriented" code. oo can be "work-averse"
#pdc "19 times out of 20 you're wrong. if you're really good" re: perf problems
#pdc "even army bridges have limits" don't plan for infinite capacity
#pdc "premature optimization is the root..." != "perf is bad do it at the end"