<asheesh>
Sorry that we didn't have a smooth answer in the install script for your situations, btw. I'll reread this on Mon and see if there's something smarter we should do.
<asheesh>
I guess there's nothing wrong with HTTPS on port 6080. It just never occurred to me.
<pod>
boom. have things configured again. ethercalc is unpacking
<asheesh>
BTW consider also Rocket.Chat within your Sandstorm (-:
<pod>
we're trying out rocket.chat for NB at the moment.
<pod>
at chat.noisebridge.net, but it's not in sandstorm.
dwrensha has joined #sandstorm
<pod>
though maybe a NB sandstorm instance w/ chat and other things would be the most useful way to do that.
<asheesh>
That is one thing that could well be true.
<asheesh>
Which is why I bring it up.
<asheesh>
I sort of hate to make the decision matrix even larger than it otherwise would be, so feel free to decide ENOBRAINSPACE.
<pod>
is there a "let new users sign up" option in sandstorm?
<pod>
that doesn't rely on 3rd party oauth services?
<asheesh>
There is email login, where we email people a link to log in with.
<asheesh>
You have to set up working outbound SMTP on the Sandstorm host for that to work, FWIW.
<asheesh>
A cool thing (?) is that you can let Google-loving people use Google login, and let Google-non-loving people use email login, and hopefully everyone's happy.
<pod>
that might work. as long as there's a non-google option
<pod>
for $CULTURE reasons
<asheesh>
Ya for sure.
<asheesh>
+1 to your $CULTURE reasons
<pod>
can't have your anarchist hackerspace rely on the GOOG
<asheesh>
Back In The Day, I was a founding member of Noisebridge, and I remain ++ on the enterprise
<pod>
your name does pop up over the place :)
<pod>
didn't realise you went that far back. neato !
<pod>
SF of that time period sounds like a very interesting place to have been.
<mnutt>
is there any point in posting my app here and asking people to try it out before submitting to the app marketplace, or are the same people going to see it either way?
<asheesh>
IMHO sandstorm-dev is best place for informal feedback, second-best is IRC
<asheesh>
App market is not super great for informal feedback because Kenton will just click Yes or No
<asheesh>
We should document this presumably
<asheesh>
Then again maybe Kenton should just click Yes and then people can try it out via app market and then you can get feedback that way.
<dwrensha>
mnutt: one reason not to post to the marketplace might be if you thought there was a chance you'd need to change the app in an incompatible way
<mnutt>
got it. I don’t believe I’ll need to, but I’ll post to the dev list first
<kentonv>
yeah I basically only look for typos or misunderstood metadata fields when approving things for the app market
<kentonv>
I actually don't even have a way to test-install the app from the app market interface... need to fix that
ArcTanSusan has quit [Quit: ArcTanSusan]
paroneayea has quit [Read error: Connection reset by peer]
mnutt has quit [Quit: mnutt]
rustyrazorblade has quit [Quit: rustyrazorblade]
rhapsodhy has quit [Remote host closed the connection]
rhapsodhy has joined #sandstorm
indiebio has joined #sandstorm
ArcTanSusan has joined #sandstorm
ArcTanSusan has quit [Quit: ArcTanSusan]
jadewang has quit [Remote host closed the connection]
PMT has joined #sandstorm
neynah has joined #sandstorm
ecloud_wfh is now known as ecloud
dwrensha has quit [Ping timeout: 244 seconds]
kentonv has quit [Ping timeout: 250 seconds]
aaronr has quit [Remote host closed the connection]
paroneayea has joined #sandstorm
XgF has quit [Ping timeout: 255 seconds]
XgF has joined #sandstorm
mnutt has joined #sandstorm
rolig has quit [Ping timeout: 260 seconds]
rolig has joined #sandstorm
natea has joined #sandstorm
aldeka_ has joined #sandstorm
snolahc1 has joined #sandstorm
micahd_ has joined #sandstorm
garrison_ has joined #sandstorm
todayman_ has joined #sandstorm
todayman has quit [*.net *.split]
aldeka has quit [*.net *.split]
augustl has quit [*.net *.split]
oao has quit [*.net *.split]
micahd has quit [*.net *.split]
KaZeR has quit [*.net *.split]
snolahc has quit [*.net *.split]
garrison has quit [*.net *.split]
micahd_ is now known as micahd
KaZeR has joined #sandstorm
mnutt has quit [Quit: mnutt]
augustl has joined #sandstorm
mnutt has joined #sandstorm
jjb has joined #sandstorm
ArcTanSusan has joined #sandstorm
ArcTanSusan has quit [Quit: ArcTanSusan]
bb010g has quit [Quit: Connection closed for inactivity]
funwhilelost has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
mnutt has joined #sandstorm
<asheesh>
mnutt: Pseudo-random question for you: Do you have any interest in possibly supporting Static Publishing with davros?
<asheesh>
I ask because I am currently attempting to mass-package all the Meteor apps I can find on GitHub and I find I need a place to store the SPKs publicly and it seems like davros would be a perfect way to send the SPKs up to a server, but then I have the problem of how to cause those files to be available on the web.
<asheesh>
I guess for now I could do a hack where I nginx-expose the grain's directory if I'm doing this on a server that has nginx as well as Sandstorm.
<asheesh>
So maybe I should live with that hack for now.
<mnutt>
asheesh: yeah! I was trying to think of the best way to do it. does sandstorm support hard links within grains?
<mnutt>
that way you wouldn’t end up doubling your storage space
<asheesh>
Ohhhh
<asheesh>
I think so re: hardlinks
<asheesh>
I don't see why not
<kentonv>
symlinks would probably be better
<mnutt>
will static publishing work ok with symlinks?
<kentonv>
hardlinks work fine, but when you try to download a backup, zip is going to include two copies of everything, most likely
<mnutt>
(I guess I don’t see why not, if it works with hard links)
<kentonv>
yep
<kentonv>
oh interesting, hmm, you'll need to use relative symlinks
<mnutt>
that’s doable
<kentonv>
it looks like the "serve www file" code path actually executes in the supervisor's directory (which is one step up from the sandbox), which could maybe be a security problem some day (not currently since we don't yet store anything interesting there)
<kentonv>
we probably need to manually interpret symlinks. Ugh.
<ocdtrekkie>
asheesh: Goal of autopackaging Meteor apps en masse? To show those devs?
<ocdtrekkie>
b/c obv. Many mass packaged apps won't be tailored well to Sandstorm out of the box.
funwhilelost has joined #sandstorm
<maurer>
kentonv: If I have an empty list, does the target of the pointer to the first element of the list canonically have to be null?
<maurer>
kentonv: This seems "obvious", but I can't find anything in the text of the encoding specification to mandate it
<kentonv>
what do you mean by the target being null? Do you mean does the pointer have to be null? (which means it has no target)
<kentonv>
assuming that's what you mean: no, a null pointer and a pointer to a zero-sized list are distinguishable. The hasFoo() method will tell the difference.
<maurer>
kentonv: What I mean is in the encoding of a list, if the size field for the non-composite case (e.g. field D) is 0, does B need to be zero
<kentonv>
no
<maurer>
If it is allowed to vary, than can't we have multiple encodings of the same message?
<maurer>
and suddenly canonicity doesn't work so hot?
<maurer>
e.g. I can have 2^30 different encodings of the empty list for any given type
<kentonv>
it's a pointer to a location. The same canonicalization rules apply: the pointer should point in between the objects before and after
<maurer>
OK, so it has a fixed value, and the value is _not_ null
<maurer>
got it
<kentonv>
a null pointer always has A = 0
<kentonv>
so if A = 1 (list), then it's not null, and has to point to a location
<kentonv>
a zero-sized location
<kentonv>
so the location will actually be the same as the next pointer in order
<maurer>
Yeah, makes sense, that's actually easier for me to implement too :)
<kentonv>
I just realized a weird corner case
<kentonv>
so if you have a non-null struct pointer to an empty struct, then B cannot be zero, because zero would make the pointer null
<maurer>
kentonv: B being zero would overlap the struct pointer itself though, yeah?
<kentonv>
if canonicalization would otherwise make B = 0, then you should instead set B = -1. This is an arbitrary convention that already exists in the code.
<kentonv>
no because it counts from the end of the pointer
<maurer>
Oh, OK
<kentonv>
so B = -1 actually points back to the pointer itself
<ocdtrekkie>
I think I topped HN. And then my submission got buried. :P
<kentonv>
ocdtrekkie: ?
<ocdtrekkie>
I submitted the best news of the day, it got 194 upvotes.
<ocdtrekkie>
But isn't on the front page. :P
<ocdtrekkie>
IMHO Star Trek matters to tech news. Since all our modern inspiration came from it.
<ocdtrekkie>
(But I'm not shocked an HN mod decided it shouldn't be on the front page.)
paroneayea has joined #sandstorm
decipherstatic has joined #sandstorm
paroneayea has quit [Quit: b]
paroneayea has joined #sandstorm
<maurer>
kentonv: Sorry to keep asking dumb questions, but if a canonical composite list has length 0, does its struct tag necessarily have data and pointer size sero?
<maurer>
*zero
<maurer>
(code seems to indicate that the tag is expected to always exist, even if the list has length 0, but since the code doesn't normally access a length 0 tag, and I don't see a reference in the spec, I don't know what the data_size/ptr_count fields are supposed to be in the tag if the list is empty)
<kentonv>
maurer: Normally, the struct segment sizes need to be the minimum possible that avoids dropping data from the elements. Since there are no elements, yeah, I'd say they should be zero.
<kentonv>
(that is, for canonicalization purposes they need to be reduced to the minimum possible)
<maurer>
OK. Zero just seemed like a funny case because with no elements, there is no data to do a truncation check on
<kentonv>
I agree it's more like "NaN", but let's say the rule is to make it zero. :)
simonv3 has joined #sandstorm
natea has quit [Quit: natea]
natea has joined #sandstorm
<zarvox>
mnutt: I'm trying to use the owncloud client as packaged in Fedora 22 (mirall-1.7.1) with Davros on Oasis, and the owncloud client keeps hitting either "CSync fatal parameter error. 403 Forbidden" or "An internal error number 1060 happened. Could not read chunk size: Secure connection truncated".
<mnutt>
zarvox: ahh, I’ve only really done much testing in mirall 2.x. It’s possible that they changed the authentication code. I can try 1.7 in a VM later tonight.
<zarvox>
Maybe the latter suggests that our nginx config also needs adaptation to support the owncloud client, or something? Maybe the client and server can't agree on a ciphersuite or something?
<zarvox>
Alternately, I can wait like a week for Fedora 23 to be out :D
<mnutt>
I’m assuming nginx just passes through everything to sandstorm, right?
<mnutt>
it’d need to pass all headers along or things are going to go badly
<zarvox>
That's the default, with a few headers (X-Real-IP, for instance) being overridden, but only those specific things.
jadewang_ has joined #sandstorm
jadewang has quit [Read error: Connection reset by peer]
<zarvox>
oh, whoa, there's both the "owncloud client" and "mirall"
<zarvox>
I shall try the former instead!
<asheesh>
Debian seems to only have "owncloud-client" not "mirall" so I went with the former and got an errorl.
<asheesh>
+/- l
<zarvox>
I am having success with Fedora 22's "owncloud-client" package!
<zarvox>
owncloud-client-2.0.1, FWIW
<maurer>
dwrensha: https://github.com/maurer/capnproto-rust/tree/wip/canonical I need to write some tests for this now (I'm sure I've gotten something wrong), but does this look like an appropriate way to deal with your abstraction structure/api?
<maurer>
dwrensha: (I remember you sounding worried when you realized I was going to have to poke private/layout.rs, so I figured I'd point you at this draft as I begin trying to test it)
<dwrensha>
maurer: cool, I'll try to take a look later today
<zarvox>
mnutt: the web UI appears to show a .jsdav file in a folder I created from the owncloud client; is that something you intend to show?
<zarvox>
I'm guessing it's a davros implementation detail, since you use the jsdav library?
<asheesh>
I guess I have 2.0.0+dfsg-1 and it tells me: "Failed to connect to the secure server addres https://sandstorm@api.mnutt.sandcats.io/. How do you wish to proceed?" The two options are: "Select a different URL" and "Retry unencrypted over HTTP (insecure)"
<asheesh>
fwiw.
<asheesh>
I'm cool with being told to upgrade, and/or I can try to tell you what exactly this code is doing if that's useful.
<maurer>
So, I see in several security advisories references to using AFL to test capnproto, but do not see any of the scripts for doing those tests in the capnproto repo - are they elsewhere?
<maurer>
(I suspect that dwrensha's set_root() on a new message builder with sufficiently sized segment is almost a canonicalization routine, and want to check that intuition)
NOTevil has quit [Quit: ....]
<kentonv>
maurer: Ben Laurie ran it. I don't recall him submitting the scripts.
zeroish has quit [Read error: Connection reset by peer]
<maurer>
I see. OK, I thought it might be a good way to see if something I thought generated canonical versions of messages actually did, but I guess I'll just try to write some tests and hope it's enough
funwhilelost has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
geofft has joined #sandstorm
funwhilelost has joined #sandstorm
decipherstatic has quit [Remote host closed the connection]
zeroish has joined #sandstorm
funwhilelost has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
simonv3 has quit [Quit: Connection closed for inactivity]
mnutt has quit [Quit: mnutt]
larjona has quit [Remote host closed the connection]
funwhilelost has joined #sandstorm
mnutt has joined #sandstorm
natea has quit [Quit: natea]
neynah has joined #sandstorm
jadewang_ has quit [Remote host closed the connection]