**** BEGIN LOGGING AT Wed Oct 11 03:00:02 2017 Oct 11 03:04:33 DocScrutinizer05, you must be a stupid arrogant teenager who got an "openmoko engineer" cloak by virtue of not knowing the first thing about mobile networking Oct 11 03:04:37 don't you even feel shame? Oct 11 03:05:06 * DocScrutinizer05 blushes Oct 11 03:05:13 sory dear Sir Oct 11 03:07:14 DocScrutinizer05, anyway, i like these two ancient greek telegraph ideas: one had two people on hills sit there with water clocks that had faces where each letter of the alphabet was printed. the person who wanted to send a message would uncover a lamp, and start the clock. the receiver, upon seeing the light, would also start his clock. when the clock hand reached the desired letter, the send would cover his lamp again, and both would stop the clock, and see Oct 11 03:07:14 the letter that was intended for sending Oct 11 03:07:41 btw what would you think about bundling (sales) smartphones with a VPN service? Oct 11 03:08:14 i don't like the concept of VPNs, it's just centralization again to me with the assumption that the centralizer is more of a good guy than the potential MITM would otherwise be :( Oct 11 03:08:54 i like VPNs like Tinc or the way Syncthing works (although that's not a VPN, but imagine it were) Oct 11 03:08:55 indeed, it is Oct 11 03:09:26 at the end of the day my solution to everything tends to be a DHT :P Oct 11 03:10:15 in the VPN case a DHT would just let you not lose track of the addresses of the other clients in your own VPN, and help negotiate NAT punching Oct 11 03:10:46 of course it's not like the *data* would go over it (although Syncthing has nodes that actually do that, when it's the only way to bypass NAT) Oct 11 03:11:33 anyway, the other telegraph idea was less technical but probably a lot more feasible: the sender just had two lamps with various different lampshades available for them, specifically, 5 per lamp (imagine: fully open, fully closed, vertical, horizontal, cross-shaped) Oct 11 03:12:06 there are 24 letters in the greek alphabet, so with a 5x5, the sender can easily send arbitrary letters by changing the lampshades in succession Oct 11 03:14:30 i think it is not quite understood whether these ideas were just musings by thinkers, or were actually put into practice during wars Oct 11 03:14:56 i bet the water clock idea wasn't put into practice successfully, though. it is fancy but just... no. Oct 11 03:17:56 if i really were a teenager... maybe i'd be smart enough not only to come up with fancy networking idea, but also smart enough to implement them, and perhaps even energetic enough to push for their adoption Oct 11 03:18:10 or not. i probably wouldn't. but still, i'd feel less powerless by a tad. Oct 11 03:20:37 DocScrutinizer05, hey, this little computer-like thing might make you smile. i have it, only crowd-anything project i ever dared back. https://www.youtube.com/watch?v=OzqcfQo7rhs&list=PLUHYzV0mEx6G46uJVHPa5TLFpxWdxp6kt Oct 11 03:22:09 (but well it failed miserably in having any degree of success, so maybe it's not so smile-worthy) Oct 11 06:35:17 DocScrutinizer05: it completely slipped my mind. (GC). I'll handle right now Oct 11 09:52:12 i wonder why linux has separate command 'swapon/swapoff' instead of just (u)mount -t swap Oct 11 09:52:34 kinda makes it unnecessary Oct 11 11:02:05 KotCzarny: because swap has pretty much nothing to do with mounts. Oct 11 11:02:53 KotCzarny: except that they both can be handled by `mount` when they appear in /etc/fstab, but that's purely a userspace oddity. Oct 11 11:02:57 yes, still is in the same domain (disk partition/block device) Oct 11 11:05:49 What about reading/writing a partition table? That has to do with disks. Should that functionality be included in `mount(1)`? Oct 11 11:06:29 thats more into mkfs/mkswap Oct 11 11:06:39 so mkfs.partition Oct 11 11:06:43 or mkfs.gpt Oct 11 11:07:04 When `mount(1)` adds swap, it doesn't even use the `mount(2)` call. Oct 11 11:07:19 and swap is independent of things like mount namespaces. Oct 11 11:07:40 The only reason it's included in `mount` is for the sake of fstab. Oct 11 11:08:11 are there other things that are in fstab but not in mount interest? Oct 11 11:08:18 so you can put the adding of swap into the same place as the mounting of filesystems. Oct 11 11:08:37 and swap is a filesystem Oct 11 11:08:46 It's not a filesystem. Oct 11 11:08:49 because it has structure Oct 11 11:08:59 not of files, but blocks Oct 11 11:09:04 well, allocations Oct 11 11:09:18 You can't mount it with `mount(2)`, it doesn't appear in the VFS, it has no filesystem structure. Oct 11 11:09:36 Otherwise you might as well call any data format a filesystem. Oct 11 11:09:40 then it might be interesting to add it to mount(2) Oct 11 11:10:27 It doesn't appear in /proc/mounts Oct 11 11:10:35 same as above Oct 11 11:11:20 there was time when there wasnt ntfs in /proc/mounts (because it wasnt implemented) Oct 11 11:11:34 If you could mount it, it would not be done to enable it as extra space for pages to swap to. Oct 11 11:11:54 It would only be done to get a representation of its contents. Oct 11 11:12:24 /proc/mounts shows your currently mounted filesystems. Oct 11 11:12:46 wouldnt it be nice if it showed used/total/free pages in df? Oct 11 11:13:04 If you could mount ntfs somehow (eg, using fuse), it should still have appeared in /proc/mounts Oct 11 11:13:45 What information would it actually show in `df`? If you reboot the system, do the numbers get reset? Oct 11 11:14:03 yes, because it would be representation of the mounted state Oct 11 11:14:16 not 'on disk' Oct 11 11:14:23 That would be a very strange usecase for `df`, and if it didn't behave that way, it would be useless because it would pretty much always appear to be full. Oct 11 11:14:36 yup Oct 11 11:16:03 It would also require special behaviour from `df`, since `df` calls `stat(2)` on the mount points it knows about (probably by reading /proc/mounts) to find information. Oct 11 11:16:31 yes, but it's up to kernel 'fs' driver to update those Oct 11 11:16:51 and again, swap is completely independent of things like mount namespaces. Oct 11 11:16:51 and you can do whatever you need Oct 11 11:16:56 similar to what fuse does Oct 11 11:17:23 If you enter a new mount namespace, you can forget about all of the mounts associated with the parent, but you'll still have the same swap. Oct 11 11:17:59 on the mountpoint you could have files similar to those in proc or sys Oct 11 11:18:24 which would allow to do config/stat type of actions Oct 11 11:18:46 But what if you're in a new mount namespace and you unmount the swap mountpoint? Oct 11 11:19:13 You won't see the swap that you will be potentially using. Oct 11 11:19:49 also, what if you don't want to spam your mount namespace pointlessly with swap mountpoints? Oct 11 11:20:18 and what happens if someone creates a new mount namespace, adds the swap, then exits? Oct 11 11:20:45 there will always be people who want/dont want particular ideas Oct 11 11:20:49 Is anyone able to turn that swap off? Oct 11 11:21:02 sure Oct 11 11:22:06 because once you mount-over, it's up to you to know the consequences and act accordingly Oct 11 11:22:31 I'm not talking about mounting over anything. Oct 11 11:22:45 I'm talking about adding swap in a private namespace. Oct 11 11:23:14 Swap isn't private (at least not in Linux); any pages can be moved into swap. Oct 11 11:23:46 and now there is a new idea, of private swapspaces Oct 11 11:24:43 and it'd be kind of crappy if you can add swap in a private namespace then exit (causing the namespace to disappear and thus the mount point) then not be able to "unmount" the swap, ever. Oct 11 11:25:08 unless you make it umount automatically Oct 11 11:25:12 private swapspaces sounds plausible, though it's obviously not the same as a mount namespace. Oct 11 11:26:01 but it would have strange behaviour sometimes. Oct 11 11:26:49 eg, if processes across different swap spaces are sharing memory, whose swap space is used when those pages are pushed out of RAM? Oct 11 11:27:15 i would guess private one would have priority Oct 11 11:27:30 They could both be private. Oct 11 11:28:15 though saying "the private one" is missing the point of namespaces. Oct 11 11:28:28 then the one that is accessible by both Oct 11 11:28:33 the point of namespaces is to behave as if everything is private. Oct 11 11:29:10 When the kernel runs the initial process, it's run in a particular mount/network/ipc/user namespace. Oct 11 11:29:54 When you `unshare(2)`, all you're doing is creating more of these namespaces. There isn't a namespace that is particularly "private". Oct 11 11:31:05 The reason you'd call it private is because not many other processes are in that namespace. Oct 11 11:39:43 what the ... ??? my tab nick-completion broke Oct 11 11:40:05 sicelo: Generalantilles in online Oct 11 11:41:28 (tab) duh, there's an option "allow tabulations" in konversation that I never seen nor used before Oct 11 11:42:27 and what's most funny: it's in input box context menu but it's channel/tab specific Oct 11 14:55:43 KotCzarny, not everything that has some kind of structure is a filesystem, or RAM itself would be a filesystem as it mirrors swap's structure, but calling RAM a filesystem is far from standard nomenclature Oct 11 15:40:15 just wait when nvm flash becomes standard Oct 11 20:39:20 DocScrutinizer05, i'm almost more worried about the audience's views expressed (including by votes) than by the corporate idea itself https://www.reddit.com/r/firefox/comments/75okqd/mozilla_might_offer_freemium_services_in_the/ Oct 11 22:05:29 hi **** ENDING LOGGING AT Thu Oct 12 03:00:01 2017