Again you misunderstand me. You keep thinking I am this stupid turd.
Well, I am! But only in Anton land ;-).
:) <3. :P lol puke.
Anton Aylward
On 10/14/2015 06:13 AM, Xen wrote:
Once again history and experience trumps your view of things! Having separate 'spindles' for different parts of the storage offers a good degree of parallelism for a number of reasons.
Duh what do you think I am. Not that I've always *had* separate drives, but my current computer at home (did) have 2 separate (SATA/IDE) disks. And I prefer, or did prefer at the time, to keep data on an external SATA drive and just backed up to an identical size USB drive. Windows may try to store all user data on the same system drive. But essentially what I've always done and many I think is to relocate that data (and you can easily now, with those libraries) to a separate partition. I have always favoured a separate data partition, even if it was on the same drive. So there was always at minimum a C: and a D: in my system, and, when it was easier to achieve, also a subst for E: and F: and so on for various "sub mounts" or sub partitions. But the main point of Linux and the problems it has with /home is that /home is not only data, but also configuration. So you won't have a user if you put it on an external drive and then disconnect that: that was the point. I was talking about USB and SD cards only, really. What I do in Windows, well, I'm still not "out" with it. eruit zijn, to have made up your mind. The problem is that there are at least three levels of storage. There is personal information (the self), collective information (the group) and universal information (the all). At the same time, there is a division in creative and consumptive. Consumptive information is usually universal, or, the other way around, universal information is almost always consumptive (it is those movies and music made by others). Universal information is typically something you want to share among the people on your computer/network/systems. Personal information you typically want to keep private. Collective information can be projects you are working on; and the information you have researched for these projects. This information is not personal because it results from research, anyone could apply it. Nevertheless it could be more or less private. So you have at least three types of storage: - personal information, secrets, agenda, contact details, phone numbers, letters you've written - projects and creative pursuits - stuff you download Ideally what you want is for there to be a way that you can layer these things on top of each other if they need be. Because now you have like a 3x5 matrix in a certain way. "video", for instance, can be in various stages. You can have personal video that you have not shared yet with anyone. Microsoft allows for this by adding various locations to their libraries. I have typically always used, because that was the easiest, the system C: drive user folders as a storage space for un-managed or un-archived or un-organised files that I create on the fly, so to speak. I'm not really good at cleaning up or tidying my home, so to speak, I have always let things get to be a mess. But once in a while I would set a few hours apart and clean up the mess and move everything to "persistent" storage in the sense of an organized tree of data, and often times I would spend a lot of time trying to reorganize that. Lately I have had --- censor ---. So I've had like those multiple levels, but projects fell in between, I did not have a dedicated location for that. I just kept them with my personal files, and sometimes stored them on the big data when they got bigger or whatever. So you could also have three levels of pictures: those you shoot yourself, or want to keep private, those that are shared among peers, and those you just download off the web. In a Linux system that would be: accessible by owner, accessible by group, and accessible by all system accounts. That is basically the idea. But it is rather hard to accomplish when there is not a thought-out plan for it yet. It really needs something of development to get something nice that way. For instance, you could have a network share from a network harddisk device that is just public for reading. That contains your collections. Then on the same n-a-s you could have stuff shared with certain users. Then your private stuff could also be on the NAS or you could keep it locally and sync it elsewhere. But how are you going to access these various layers? Are you going to make a home user "videos" folder that contains "me", "us" and "all" ? Are you going to let Microsoft show you an overview of all these sources? I guess that works... well. Enough. For now. Maybe you want a system where there is another dimension apart from a hierarchy; you want a switch that changes the context and changes the view of that folder accordingly. It would be a virtual folder sourcing from different places. Nevertheless this only pertains to information, not configuration. Configuration is that which on Linux typically resides in dot files. Now what you *can* do is .... use an overlay filesystem to combine external or remote or different volumes into the /home, not mounted on top of it, but rather perhaps together with it. I only know about aufs, but it is not available in SuSE. (OverlayFS doesn't work). On the other hand... well. Caches for instance, .cache and all that. They are rather sensitive information in a certain way so usually you'd really want them to be eCryptFS-sed and all that. But .mozilla contains firefox data as well and .thunderbird I think contains your mail store. That's not configuration but data (roaming data on Windows). Caches are in "local" on Windows. So what you gonna do. Cache fits perfectly on a home / root system, so you can just make sure all default folders are created on the root fs. Then you make sure your documents/pictures/videos etc point to another destination. Or are overlayed with aufs for example. Then you also move .thunderbird or whatever to the other source, or perhaps change thunderbird config to do that, but a symlink might do fine. [[ With aufs it is very easy to get a default setting of: folders are created on one device, but a folder that is available from another device will kept (be kept) updated on that other device/volume. So anything you move to that other device (you need to mount the sources first somewhere, which I use /mounts/ for these days) will stay on that other device and not be written to the default device. ]] It is all rather difficult. For instance when I tried to mount /usr to /mounts/user and then bind mount it to /usr, my system would not boot (Ubuntu) or at least it would not find /usr in time. In Windows I don't deal with complexities; I just arrange drive letters and volumes tied to them. Currently I am trying to discover how I can manually/automatically can keep my personal files synced across a cloud. OwnCloud is not really good yet, particularly the user interface could need destruction and more destruction and then burning it and maybe finally a phoenix will arise from the ashes. I have never seen such a bad music player, for example. A local VP... provider is building their own OwnCloud solution that they will sell (provide as a service) most likely based on OwnCloud but with their own user interface that they design. Because they kinda agreed with me that the base default user interfase is rather lacking. I don't know what Synology is doing and How I can really use that. I can hardly install synology software on some VP..x.
Yes, having huge drives means that most people using PCs for desktop think it simpler to have one drive; many don't even partition that. Certainly Windows did not have a heritage of putting user data on a different spindle or partition!
Most of my friends probably did that though. There is really nothing much sitting in the way of that. Now, you can change the base location of the libraries. For instance, my pictures etc. now point to d:\home\pictures, for now. This allows me to create smaller system images for backup and restoration of the system bare. But, and this was the point with /home, my user is also in that system image and I would not really want it to be outside of that, because user configuration is also configuration. That's why I go to pains with that "aufs" and try to find ways to keep configuration data in /home on the rootfs, but all other data (excluding caches, perhaps) on another sources/volumes/devices. In other words, I don't *just* want /home to be somewhere else. That wouldn't work, because you don't have a system without /home. Except for /root, that is not on that volume, if you were to separate it. /home is not really a very good candidate for externalising it. So I create /data or /store (still fiddling with that) and put my files on there. And I externalize (so to speak) THAT. But it is troublesome so if I don't have time for all that, I still do /home on a separate volume. To my great pain, in a way. I recently forgot to backup /home because I didn't remember it was on a separate thing. Not intuitive to me.... and stupidity of LVM perhaps.
Certainly back in the PDP/11 days, having a separate spindle for Swap speeded up the roll-in/roll-out. And having a separate spindle for /tmp speeded up many operations, since memory limitations meant that in-core caching of intermediate files and sort buffers wasn't practical. While some of that is not an issue with today's virtual memory and larger internal memory, it does raise issues of where the best position for expensive fast store such as SSD should be. As ever, there is a trade-off between memory and its speed/cost and 'external' storage and its speed/cost.
Yeah well, you don't really have to explain the basics like that. No one in his right mind would use Linux if he didn't understand the basics. Except for the poor people who get directed that way by others.
Separating code and data has always been a good principle. That is one reason we have a separate /home branch at all. Having it as a separate partition can make upgrades and backups simpler as well. Having it as a separate 'spindle' (aka 'drive' aka 'device') also offers a degree of parallels - efficiency and performance - as well as maintainability.
That is pretty much a "duh" for me. But you forget that home combines config and data. It is also very difficult for me to contain and locate data from sources such as wikis and web interfaces. Do you have any info on that? A best practice or whatever? Stuff under /var/lib is usually root writable. Or owned by www-data or whatever. It creates problems. DokuWiki (something I use) can run into great big permission problems that you can't solve. Unless you reinstall it. OwnCloud puts information under /var/lib/owncloud or whatever or even /usr/share/owncloud. And by default it does not want you to write to its data store by external actions. Which is quite a problem because its ability to import data is extremely limited. And owncloud can be (and is) a nice interface to a data store you manage yourself. So you want the data store to be within your own control, and then give that cloud solution ACCESS TO THAT. AND NOT BE THE ONLY OWNER OF IT. The funky thing about owncloud currently is that although it says you CANNOT or MAY not access the data directly, and if you do, its permissions get fcked up, eventually it automatically fixes those permissions. At least the permissions in its own database. Don't know about file (filesystem) permissions yet. It is a trouble. I want it to create files that are owned by the www-data GROUP and writable by that group but owned by a different user ideally, or for me to be able to easily convert it to my own user (without needing to be root). And I never know how to do that, because options or solutions are never available in a GUI, and I have limited time (or mostly, also, limited energy). For what you know, I might not be the most healthy person on earth, you know. And most people aren't, who like things to be easy and are not superhumans who can accomplish any seeming impossible or hard goal. So anyway I will just be symlinking owncloud locations to my own data storage so that owncloud only maintains backups, or caches, or metadata, or all that stuff. But it is terribly difficult that I have to do everything myself. And these people do not think about this stuff in advance. So you are always fighting your battles just alone, very little cooperation from others, because there appear to be no like-minded people. "Where two or three are gathered in my name." But if you are all alone. Well. Yeah. I cry. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org