@milessabin Any is what Spark is passing around internally.
-
-
Replying to @heathercmiller
@heathercmiller Gotcha ... that's the mistake IMO.1 reply 0 retweets 0 likes -
Replying to @milessabin
@milessabin Yep, that's been my argument for about a year now. Systems folks aren't used to languages with powerful/flexible type systems.1 reply 5 retweets 6 likes -
Replying to @heathercmiller
@heathercmiller how much effort do you think a PoC spike reworking those internals be?1 reply 0 retweets 0 likes -
Replying to @milessabin
@milessabin Quite an effort. Spark's a big system. I hacked around inside it for a few months & still dont fully understand some parts of it3 replies 0 retweets 0 likes -
Replying to @heathercmiller
@milessabin Trouble is that a lot of internal stuff has to be serialized & sent around too. So your serializer has to work for that case too2 replies 0 retweets 0 likes -
Replying to @heathercmiller
@milessabin It's not impossible. It's just a lot of work. I grossly underestimated it the first time I tried.1 reply 0 retweets 1 like -
Replying to @heathercmiller
@heathercmiller@milessabin The problem is that serialization is treated as a magic lower layer when in fact it should be explicit a priori.2 replies 1 retweet 2 likes -
Replying to @rolandkuhn
@rolandkuhn@milessabin Yes, I partially agree. More precisely, I think the problem is the combination of objects + auto-magic.1 reply 1 retweet 4 likes -
Replying to @heathercmiller
@rolandkuhn@milessabin I think the greater culprit of the two is the normalness of psychotically evolving object graphs in JVM-land.2 replies 1 retweet 4 likes
@heathercmiller @rolandkuhn @milessabin "Psychotically" is definitely an underused adverb in CS.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.