#dat

erikg: I like how you tried to use shell chars, but I think > for forks and write to files at the same time a bit confusing

and overall I'm a bit on the fence to what is better, | and > chars, or pipe and fork words

erikg: it's seems the chars are a bit more confusing/noisy, but maybe that's just a matter of getting used to that style. I don't like that Makefiles use a lot of weirds chars and at some point start looking more like perl regex. I think sometimes keywords are better, like in Dockerfiles

erikg ogd mafintosh: nice BAM visualisations with d3 and node.js published in nature methods: http://doi.org/10.1038/nmeth.3174

at 1:20 in the video they mention pipelines

yeah, "in the future"

what's bam?

a file format ? bmpvieira

karissa: yes, it's the binary version of SAM (Sequence Alignment/Map)

okay

cool

do you know anyone who works on it?

it's the most used format you get after you've aligned your sequences to a reference genome for example

it's used a lot in bioinformatics

cool. by *it* I mean iobio

nop

from github seems to be two dev

cool. maybe they could use dat/datscript for some pipelining stuff

bmpvieira: karissa: I am a member of this lab (iobio)

I've been trying to get them on the same page with bionode (and dat) but need to make a more concerted effort. Aaron Quinlan is also an old member.

It's the group I just left

bmpvieira: > means fork to file and >@ means fork to stream

erikg: yes, but visually I think it gets blurry

erikg: ahh, I just saw Uni. Utah and missed Marthlab

erikg: bionode used in cyREST https://twitter.com/cytoscape/status/5174669686...

almost missed that tweet

cool. erikg: it would be cool maybe to set up an example of using dat as the backend/serve point for their viz tool

and then later down the road (once we have dat script at some sort of alpha) using dat script to pipe some data in/out perhaps & show the power of both together.

will be in Oakland from 12/7 - 12/14

[dat] podviaznikov opened pull request #224: Fix example. Should be an array (master...patch-2) http://git.io/f5Clhw

karissa: it should be easy for them to work from dat, provided indexes on quantities of interest

karissa: also the iobio demo has been useful from time to time for randomly QC-ing alignments, but for some reason they seem to be stuck in demo land. grant applications.

:/

pkafei: w00t

erikg: https://www.cryptocoinsnews.com/make-backup-gen...

ok, i guess i should do this huh?

mine fits in just 400kb... is that indicative of some kind of health issues?

bmpvieira: i like their style

erikg: me too

erikg: 400kb, maybe you need to update your genome or something :P

bmpvieira: this is totally what i want to happen, but i think they're being a bit silly about the 750mb stuff. but the hype is good.

bmpvieira: i have a mid-80s model

erikg: yeah you're probably missing the latest features, and I've heard now it's also more "flat"

erikg: or maybe I'm confusing it with something else :P

a nice ui to model https://try.github.io/levels/1/challenges/1

karissa: try github is awesome- it has a slick interface

mafintosh: ping, im doing the live socket upgrade switching thing in my high level socket lib

ogd: i'm here

mafintosh: so i have httpsocket and im writing a new module called sockets that starts by establishing a httpsocket connection and then if a websocket can be established it should switch to that

mafintosh: so to hot-swap the stream, i should just be able to duplexify.setReadable and setWritable the new one once the old one is closed/drained etc?

mafintosh: i forget what you said on the call about a feature it might need

ogd: that would work

ogd: i don't think it needs any new features for this

ogd: the feature we were talking about was the end:false option

mafintosh: ah right

Just curious; is this for the browser?

trygve_: yep

how often do you not have a websocket these days?

trygve_: browsers have them, but we recently launched a websocket-only app and ive got a ton of reports from users saying it cant connect. probably because some http proxies dont support 'upgrade' events

we run a pretty large socket app here and running on ssl fixes that.

trygve_: i would use socket.io or sock.js or engine.io or primus or whatever but they are all so complicated

trygve_: ah yea good point

I've come to always ssl on websockets. then your next enemy is mobile devices and even worse mobile devices on the move.

trygve_: mobile devices are bad because they dont have websockets or some other reason?

mobile devices have good websocket support, though your problem will be that a mobile device on the move will have variable latenzy. Thats an issue.

then, the network will drop a bit without the browser knowing when user move between radio transmitters. then all socket connections drop (also long polling).

trygve_: ah so you are saying that tcp alone does not provide enough redundancy?

so, you do need ping/pong messages that give you information about latenzy. and you need a reconnect strategy on top of the ping pong messages.

yep

trygve_: what about sse?

xat-: sse is just http long polling so it should have the same issues he describes

sse are a bit better since the browsers have implemented reconnects.

wouldnt that be the perfect solution for mobile problems, since it sends the last transmitted id in the next request

you also have a third issue on mobile when it comes to sockets and that is that the mobile jumps into sleep mode. Ex; when the lock screen goes on the browser will silently kill all connections without informing you to save battery.

and this happens in a magical way. so you need to detect that your app gets focus and reconnect.

trygve_: ahh that is crazy

sockets to the browser is not easy. its so much more you need around it than just the socket. I've recently switched to Primus and its the first library I've seen that deal with all these issues in a nice way.

trygve_: oh good to know

I'm only using the ws module + Primus now. Pretty nice to be honest. Though; ws does seriously need a rewrite...

trygve_: i have used ws for a while but havent really looked at the internals. what are the major issues with it?

code is a bit old. einaros is not really maintaining it anymore. some of the older ws drafts can be removed since the browsers supporting those are off the market now. it lacks support for the node stream api's. stuff like that.

there are also some C bindings in there which really is not needed anymore.

Talked with Arnout about it in Amsterdam a month ago and he would love to rewrite ws, but he don't have time (me neighter).

trygve_: ah nice, i might take it on

that would be awesome. the node world needs a new websocket module (I know there has been talks about having native support in node for websockets. iow, including the ws module in the distro).

trygve_: do you happen to know if you can do wss:// from a http:// site?

bmpvieira: you going to dockercon next week?

ogd: yes you can

mafintosh: so i think we should upgrade try-dat to use wss://

mafintosh: should fix our connection issues

trygve_: whats the deal with https://www.npmjs.org/package/websocket ?

ogd: not sure. I tried it a year ago. Then it seemed a bit abandoned but seems to be picking up traction again now.

trygve_: just found this https://www.npmjs.org/package/yawl

trygve_: looks like what we want

not seen that one actually

trygve_: superjoe writes good code

ogd: a bit off topic and not node, but seen this one: https://github.com/namedwebsockets/namedwebsockets

trygve_: interesting

Rich Tibbett is a smart guy. Opera guy actually

ogd: i think i bought a ssl license for it actually and forgot about it