Command synopses should refer to "grid" rather than "virtual drive" #892

New Issue

tahoe-lafs · 2010-01-10T06:37:36Z

davidsarah commented

2010-01-10 06:37:36 +00:00

For example,
get Retrieve a file from the virtual drive.
should be
get Retrieve a file from the grid.

"Virtual drive" sounds too much like, say, a Windows drive mapping, which is not how users think of it when using the CLI.

For example, `get Retrieve a file from the virtual drive.` should be `get Retrieve a file from the grid.` "Virtual drive" sounds too much like, say, a Windows drive mapping, which is not how users think of it when using the CLI.

tahoe-lafs added the

labels 2010-01-10 06:37:36 +00:00

tahoe-lafs added this to the 1.6.0 milestone 2010-01-10 06:37:36 +00:00

davidsarah commented

2010-01-10 06:38:28 +00:00

Also there are a couple of uses of "virtual drive" in documentation.

warner commented

2010-01-10 07:48:08 +00:00

"virtual drive" should certainly go away.. much of the text which references it was written back in the era of the ambient "/private" URL, when we still had the idea of a single virual drive per user. Now that we encourage people to think in terms of a graph of dirnodes and rootcaps-as-starting-points, that's not such a great term to use.

I'm ok with "grid", but I'm not convinced it's all that great either. When I hear "grid", I think of the mesh of storage servers, or some abstract cloudy storagey thing. I don't immediately think of the graph of dirnodes and filenodes.

I guess "grid" is perfectly file for command synopses. At some point, we need to come up with a good term for the filenode graph (maybe "filenode graph" :) for use in the other pieces of documentation that refer specifically to it as opposed to the backend storage system.

"virtual drive" should certainly go away.. much of the text which references it was written back in the era of the ambient "/private" URL, when we still had the idea of a single virual drive per user. Now that we encourage people to think in terms of a graph of dirnodes and rootcaps-as-starting-points, that's not such a great term to use. I'm ok with "grid", but I'm not convinced it's all that great either. When I hear "grid", I think of the mesh of storage servers, or some abstract cloudy storagey thing. I don't immediately think of the graph of dirnodes and filenodes. I guess "grid" is perfectly file for command synopses. At some point, we need to come up with a good term for the filenode graph (maybe "filenode graph" :) for use in the other pieces of documentation that refer specifically to it as opposed to the backend storage system.

zooko commented

2010-01-10 17:32:09 +00:00

The term I've been using for that is "the decentralized filesystem": source:docs/architecture.txt?rev=4134#L17.

davidsarah commented

2010-01-10 20:30:42 +00:00

Whenever someone proposes a term I always count the number of syllables: "de-cen-tral-ized file-sys-tem" (7). "Grid" (meaning the abstract cloudy storagey thing implemented on top of a particular mesh of servers) is both much more concise, and sounds more Gibsonesque :-)

The terminology collision with "grid layer" could be handled by renaming the latter to "mesh layer". That would allow us architecture types to continue to make precise distinctions (needed for #869, for instance).

Whenever someone proposes a term I always count the number of syllables: "de-cen-tral-ized file-sys-tem" (7). "Grid" (meaning the abstract cloudy storagey thing implemented on top of a particular mesh of servers) is both much more concise, and sounds more Gibsonesque :-) The terminology collision with "grid layer" could be handled by renaming the latter to "mesh layer". That would allow us architecture types to continue to make precise distinctions (needed for #869, for instance).

davidsarah commented

2010-01-10 20:38:09 +00:00

Grepping the source tree for "grid", there are a lot of uses, but very few of them are using it in the precise sense defined in source:docs/architecture.txt

zooko commented

2010-01-11 05:06:12 +00:00

For what it is worth (and I'm not sure that is very much) "Grid" was the term for computing-as-infrastructure before "Cloud" became the term for that. The "Grid" computing people are more from a scientific computing background and the Cloud computing people are more from Net-business, I think.

I use "the grid" brevity when it doesn't matter which layer I'm talking about, as in "the put command uploads a file to the grid", but if I want to distinguish between (a) the set of machines, (b) the distributed key-value store layer, (c) the filesystem layer, then I use the longer term "decentralized filesystem".

So, what I'm saying is I welcome your ideas about terminology, and I don't think that my current terminology is perfect, but if you want to use "grid" to mean the filesystem layer, then what's the key-value-store layer? Or are they both the same for this terminology?

For what it is worth (and I'm not sure that is very much) "Grid" was the term for computing-as-infrastructure before "Cloud" became the term for that. The "Grid" computing people are more from a scientific computing background and the Cloud computing people are more from Net-business, I think. I use "the grid" brevity when it doesn't matter which layer I'm talking about, as in "the put command uploads a file to the grid", but if I want to distinguish between (a) the set of machines, (b) the distributed key-value store layer, (c) the filesystem layer, then I use the longer term "decentralized filesystem". So, what I'm saying is I welcome your ideas about terminology, and I don't think that my current terminology is perfect, but if you want to use "grid" to mean the filesystem layer, then what's the key-value-store layer? Or are they both the same for this terminology?

davidsarah commented

2010-01-11 05:42:38 +00:00

Replying to zooko:

I use "the grid" brevity when it doesn't matter which layer I'm talking about, as in "the put command uploads a file to the grid", but if I want to distinguish between (a) the set of machines, (b) the distributed key-value store layer, (c) the filesystem layer, then I use the longer term "decentralized filesystem".

So, what I'm saying is I welcome your ideas about terminology, and I don't think that my current terminology is perfect, but if you want to use "grid" to mean the filesystem layer, then what's the key-value-store layer?

I don't want to use "grid" to mean the filesystem layer; I want to use it when it doesn't matter which layer. The key-value-store layer would be renamed to "mesh layer".

Replying to [zooko](/tahoe-lafs/trac-2024-07-25/issues/892#issuecomment-74509): > I use "the grid" brevity when it doesn't matter which layer I'm talking about, as in "the put command uploads a file to the grid", but if I want to distinguish between (a) the set of machines, (b) the distributed key-value store layer, (c) the filesystem layer, then I use the longer term "decentralized filesystem". > > So, what I'm saying is I welcome your ideas about terminology, and I don't think that my current terminology is perfect, but if you want to use "grid" to mean the filesystem layer, then what's the key-value-store layer? I don't want to use "grid" to mean the filesystem layer; I want to use it when it doesn't matter which layer. The key-value-store layer would be renamed to "mesh layer".

zooko commented

2010-01-11 14:33:50 +00:00

The key-vale-store layer of Tahoe-LAFS is similar to the NoSQL systems: http://en.wikipedia.org/wiki/NoSQL . I would really like to emphasize the similarities so that people who are familiar with NoSQL (which is a rapidly growing set) will have an easier time learning about Tahoe-LAFS. How about calling the the "key-value-store layer". :-)

The key-vale-store layer of Tahoe-LAFS is similar to the NoSQL systems: <http://en.wikipedia.org/wiki/NoSQL> . I would really like to emphasize the similarities so that people who are familiar with NoSQL (which is a rapidly growing set) will have an easier time learning about Tahoe-LAFS. How about calling the the "key-value-store layer". :-)

davidsarah commented

2010-01-14 03:51:27 +00:00

Attachment goodbye-vdrive-diff.txt (47181 bytes) added

Diff to remove references to 'vdrive' and 'virtual drive', and some other cleanups to architecture.txt and command synopses

**Attachment** goodbye-vdrive-diff.txt (47181 bytes) added Diff to remove references to 'vdrive' and 'virtual drive', and some other cleanups to architecture.txt and command synopses

goodbye-vdrive-diff.txt

46 KiB

warner commented

2010-01-14 20:19:43 +00:00

reviewed and committed, changeset:d3d1293d2fee8b62. My only changes were to wrap a few long lines. Thanks!

warner added the

fixed

label 2010-01-14 20:19:43 +00:00

warner closed this issue

2010-01-14 20:19:43 +00:00

Sign in to join this conversation.