2011--improving_styx_high_latency.pdf

Notices where this attachment appears

1. Embed this notice
   pistolero :thispersondoesnotexist: (p@freespeechextremist.com)'s status on Tuesday, 30-Jan-2024 04:32:54 JST pistolero :thispersondoesnotexist:

   in reply to
   @icedquinn @a7
   
   > filing things in to a tree is just a shitty degraded form of a search index.
   
   No, no, this is like saying that you don't need to name something as long as you can point at it.
   
   Having used both a Terrible iThing and a real computer, I can say that a real computer is nicer for most tasks that I want to use a computer for. The concept of a place is not mutually exclusive with the concept of a search index. Search indexes are fine if you remember what you want. Sometimes I don't even know what I want, but I know where it is: you can't replace "walking around the neighborhood" with "type the thing you hope to find into Yelp". With named things in a place, I can communicate this with other people. (I think coders that do a tour of duty in tech support end up cured of some of their notions about UIs. You can communicate a shell one-liner much easier than you can communicate )
   
   I also can't search the entire internet: just searching the web is a big enough pain in the ass half the time. It's simple enough to index your own stuff, but indexing the system's stuff adds in a lot of cases context that you don't want and in a lot of other cases, misses things you want. Indexing the web is a large task. Indexing the entire net is another thing entirely, no one's even attempted it. You get these massive efforts like Shodan and all they do is index the type of thing that sends whatever the data is, it doesn't even have the data.
   
   > outlook mail users.
   
   I think this is the problem. That kind of study gives you data that is over-fit for the task, you can't really extrapolate. I don't need to communicate a specific thing by name to anyone if I'm looking for something in my inbox or I want to send someone something. Paths, and more generally URLs, are pass-by-reference; email is pass-by-value.
   
   Incidentally, on this topic, Octopus is still very cool and a pretty interesting thing. One of the fun parts is that it's this tiling UI mostly concerned with text, but it's also fully serializable, and that makes it possible to do playback. (mycroftiv's hubfs stuff had a chunk of that but he just used rio and acme. It was cool that he'd be able to avoid configuring a system by just interactively typing scripts and then the system got into a state he wanted and he'd just play it back on boot, so even if machines went down and came back, he barely noticed.) I used Octopus for a while; the upside is it's one of the nicest environments I've used, probably the only one that is both persistent and device-independent, the downside is that it required discarding a lot of things that you get from Inferno.
   2006--building_the_octopus.pdf
   2006--give_me_back_my_personal_mainframe.pdf
   2007--distributed_smart_spaces.pdf
   2007--octopus_as_middleware.pdf
   2007--styx_batching_for_high_latency.pdf
   2007--towards_persistent_distributed_interfaces.pdf
   2007--ui_in_octopus.pdf
   2008--octopus_user_manual_2nd_ed.pdf
   2008--using_the_octopus.pdf
   2009--batching_design_pattern.pdf
   2010--upperware_resources.pdf
   2011--improving_styx_high_latency.pdf
   2012--octopus_upperware_based_system.ps
   2012--personal_pervasive_environments.pdf
   belt.pdf