[pot@gnu.org: Re: TRAMP2 path definitions.]
Well, I tried to write a message for the Emacs developers and pretesters lists, but could not manage to write anything that would make sense. I suspect that we (ar at least it's me) do not have ideas clear enough to be explained outside of here. In fact, I do not manage to make a resume of our discussion in a way that is generally significant. Sorry.
Re: TRAMP2 path definitions.
On Tue, 13 Mar 2001, Pete Forman wrote: Daniel Pittman writes: As such, I want to propose an alternate tag to indicate our own paths: "/!/" [...] I agree with you but would like to take it further. How about reworking the whole syntax along these lines. /![@enc]/:telnet://[usr[:pwd]@]host1[:port]/ssh://host2/:/path/to/file ^^ ^^^ ^^^ ^^ ^ ^ ^^^ ^^ A B C F HG I DF E A tramp-prefix-tramp-name "/!" or "/tramp" or "/:" or "/" B tramp-prefix-encoding "@" or ":" or "/[^:]" C tramp-prefix-first-connection "/:" D tramp-prefix-next-connection "/[^/:]" or "[^/]/[^/:]" or "/[^:]" E tramp-prefix-remote-pathname same as C F tramp-prefix-authority"//:" or "#" G, H, I, etc. are specific to schemes C, D, and E are mutually constrained to match at the start. Having F set to "#" makes parsing easier for them. I suggested "//:" for F to match the syntax of URIs; see RFC 2396 and 1738. The syntax is clean enough, I guess.[1] I don't see any technical reason it couldn't be used. A couple of points, though: '//' causes problems with the default XEmacs behaviour[2] that typing '//' in a path elides it up to the second '/' and makes it an absolute path. It also conflicts with anything that emulates `substitute-in-file-name' without calling our handler. That /shouldn't/ happen, I think, but it's a risk. Secondly, I am very opposed to including passwords in the path. It's very easy for someone to C-x C-f UP UP and see the password that way. It's considerably harder for someone to strings(1) a running process or core-dump Emacs without your noticing. Not giving end users a loaded gun is, in my opinion, a nice thing to do. Passwords in the path, the history, the kill ring, the X system clipboard[3] and so on, make for an easy way to leak a password. There would be no "@encoding" clause for an out-of-band method. For in-band, encoding would be "base64" or "uuencode". We may want to add other encodings later such as "quoted-printable". I have been thinking on this for the last day or so. It's amazing what you can achieve when you don't spend all your effort writing code. I think that the idea of 'encodings' is not quite what is really wanted for tramp. scp(1) isn't a way of connecting to a remote host, it's a way of transferring data. So is base64 to mimencode(1). So, it your scheme, I think that scp /would/ fit into the first bit, like base64. The encoding might also be specified as "b64" or "uu" or "b" or "u". The protocols could also accept long and short forms, e.g. "telnet", "tn" and "t" would be equivalent. Of course. If it's not easy, it's not worth it. :) Daniel Footnotes: [1] I think this is my natural hubris emerging. I like my idea better because, um, because I /do/. :) [4] [2] I think it's the default. It's been so many years since I used a default XEmacs... *testing* Ah, yes. It is default. [3] Recent XEmacs versions put the top of kill in there by default. /I/ think this is a silly idea, but who am I to argue... [4] In other words, I don't see any real /technical/ argument against your idea. It just feels... different. Maybe not even wrong. -- Reality is not as strong as perception. Perception all too often swallows reality and spits it out in a new, unrecognizable form. -- Maytee Aspuro
Re: TRAMP2 path definitions.
On 13 Mar 2001, Stefan Monnier wrote: "Pete" == Pete Forman [EMAIL PROTECTED] writes: /![@enc]/:telnet://[usr[:pwd]@]host1[:port]/ssh://host2/:/path/to/file [...] F tramp-prefix-authority"//:" or "#" Clearly ""//:"" won't do since Emacs tends to interpret it directly. I liked the @enc thing, except that I'd use it for the file-name-coding-system rather than for the base64/uuencode. Isn't C-u [1] C-x C-f what you want there, or is that just XEmacs? It prompts me interactively for a coding system to visit with... Daniel Footnotes: [1] universal argument -- In the present-day reconstruction of physics, fragments of the Newtonian concepts are stubbornly retained. The result is to reduce modern physics to a sort of mystic chant over an unintelligible universe. -- Alfred North Whitehead, _Modes of Thought_
Re: TRAMP2 path definitions.
On Tue, 13 Mar 2001, Francesco Potorti` wrote: As such, I want to propose an alternate tag to indicate our own paths: "/!/" Seconded. I still do not like it at all. It's completely different from anything that is done out there. There is a quasi-standard for doing such things, which involves using /protocol:, and that's reasonable, widespread and people are accustomed to it. We do not want to introduce Yet Another Nice Convention. We should instead raise the problem on the internal Emacs developers lists, and then contact to Xemacs developers (or the other way around, if you care more about Xemacs), and try to find a real solution. I think this is a good idea - but even if we succeed, supporting old code is a burden. If the EFS changes made it into the up-coming 21.4 release, it's going to take a /long/ time for that to be common, let alone universal, amongst XEmacs users. Ditto the impending Emacs release, I imagine. [...] I volunteer to raise the question on Emacs lists, if I am asked to. Please do. It would be nice to see the issue addressed. I don't think it solves the problems of TRAMP if there is a better way introduced, though. I need to submit a patch for `file-remote-p' to start moving towards this better world, though. It's worth aiming for - just not a silver bullet. :) Daniel -- Sweet desert rose This memory of Eden haunts us all This desert flower, this rare perfume Is the sweet intoxication of the fall -- Sting, _Desert Rose_
Re: TRAMP2 path definitions.
On Tue, 13 Mar 2001, Edward J. Sabol wrote: I agree with Francesco. I don't like "/!/" at all. The prefix will be changeable, using custom, with no trouble at all. First off, I have to use the shift key to type the exclamation point. So, pick something different. "/tramp/" is available. ;) Secondly, there's no mnemonic for it at all. Given Emacs isn't really about that, I am unconcerned. "/tr:" was the best idea that's been mentioned. That will fix the DOS problem, though perhaps not the EFS problem, No. It's the `:' I want to avoid, specifically. I don't care if the default is "/tr/" or "/this-is-a-/really/-silly-prefix/" but we've lived with that since the inception of tramp/rcp/rssh and I don't consider it a problem personally. It's hard to support, requires hacking the innards of other packages and introduces load-order dependencies in the packages. This makes it less than pleasant to use, as the second bug report on the new Debian package shows. The poor user had not loaded the FTP access package first and then TRAMP lost. :/ Daniel -- A drug is neither moral nor immoral--it's a chemical compound. The compound itself is not a menace to society until a human being treats it as if consumption bestowed a temporary license to act like an asshole. -- Frank Zappa
Re: TRAMP2 path definitions.
On 13 Mar 2001, Stefan Monnier wrote: "Francesco" == Francesco Potorti` [EMAIL PROTECTED] writes: Emacs should have a general hook for /[^/]+:.* filenames, where [^/]+ is the protocol. Then, different packages could register to that hook and tell it which protocol they do manage. I think using /protocol: is just fine, indeed. But I don't think it requires any changes to ange-ftp or EFS. All that matters is that ange-ftp or EFS have lower precedence, i.e. that they be loaded before, As I understand it, Emacs and XEmacs both agree that changes to the overall system state are undesirable when you load a single package. While an exemption is clearly made in the case of Ange-FTP, I don't think that loading another remote access tool to work around our... issues with it is right. which is already the case of ange-ftp (since it is preloaded). Being an Emacser I don't care much about EFS. It's really not very different, to be frank. The hacks to work around it are pretty much identical. Daniel -- Sanely applied advertising could remake the world. -- Stuart Chase
Re: [pot@gnu.org: Re: TRAMP2 path definitions.]
On Wed, 14 Mar 2001, Francesco Potorti` wrote: Well, I tried to write a message for the Emacs developers and pretesters lists, but could not manage to write anything that would make sense. I am happy to offer what help I can. I would like to see a good solution to the issue forged. I suspect that we (ar at least it's me) do not have ideas clear enough to be explained outside of here. Feel free to contact me off-list if you want to work through the whys and wherefores of it. In fact, I do not manage to make a resume of our discussion in a way that is generally significant. Sorry. Huh. You have been helpful to me, at least, in my thinking. You do not give yourself enough credit. Daniel -- All that we see or seem is but a dream within a dream. -- Edgar Allen Poe, _A Dream Within a Dream_
Re: TRAMP2 path definitions.
On 15-Mar-2001, Daniel Pittman wrote: On Tue, 13 Mar 2001, Edward J. Sabol wrote: I agree with Francesco. I don't like "/!/" at all. The prefix will be changeable, using custom, with no trouble at all. Yes, that's always been the case, but I strongly feel that the default should be wisely chosen. First off, I have to use the shift key to type the exclamation point. So, pick something different. "/tramp/" is available. ;) Great, let's go with either "/tr/" or "/tramp/" then. I just think almost anything would be better than an exclamation point. Secondly, there's no mnemonic for it at all. Given Emacs isn't really about that, I am unconcerned. That's too bad. I think you should be. "/tr:" was the best idea that's been mentioned. That will fix the DOS problem, though perhaps not the EFS problem, No. It's the `:' I want to avoid, specifically. I don't care if the default is "/tr/" or "/this-is-a-/really/-silly-prefix/" I don't particularly have a problem with getting rid of the colon. but we've lived with that since the inception of tramp/rcp/rssh and I don't consider it a problem personally. It's hard to support, requires hacking the innards of other packages and introduces load-order dependencies in the packages. Correct me if I'm wrong, but it's only a problem with EFS. However, I agree that we should fix it, if we can.
check in by grossjoh: 'tramp/lisp tramp.el,2.9 ChangeLog,2.8'
Update of /services/emacs-rcp/cvsroot/tramp/lisp In directory lucy:/export/home/grossjoh/work/kai/tramp-devel/tramp/lisp Modified Files: tramp.el ChangeLog Log Message: (tramp-methods, tramp-telnet-args): New method parameter `tramp-telnet-args'. (tramp-get-telnet-args): New function. (tramp-open-connection-telnet): Use it.
check in by grossjoh: 'tramp/texi .cvsignore,2.1'
Update of /services/emacs-rcp/cvsroot/tramp/texi In directory lucy:/export/home/grossjoh/work/kai/tramp-devel/tramp/texi Modified Files: .cvsignore Log Message: Ignore *.vrs.
Telnet connection with extra args
I have now added the `tramp-telnet-args' method parameter, similar to all the other foo-args method parameters. However, I haven't tested this. Some of you wanted this. I think Harry and Ted? Could you please test this version? (New version in CVS, head branch, not stable.) kai -- Be indiscrete. Do it continuously.
Re: TRAMP2 path definitions.
On Wed, 14 Mar 2001, Edward J. Sabol wrote: On 15-Mar-2001, Daniel Pittman wrote: On Tue, 13 Mar 2001, Edward J. Sabol wrote: I agree with Francesco. I don't like "/!/" at all. The prefix will be changeable, using custom, with no trouble at all. Yes, that's always been the case, but I strongly feel that the default should be wisely chosen. Indeed. If I didn't agree with you[1], I wouldn't have asked. [...] Secondly, there's no mnemonic for it at all. Given Emacs isn't really about that, I am unconcerned. That's too bad. I think you should be. I don't, quite honestly, think that there is a real common convention about addressing remote files, save the Unix standard (just like anything else) and the Windows standard (\\server\share\file). Neither of those is really mnemonic, nor the tramp/EFS/Ange-FTP method of /host:path. It's a matter of what people are used to. [...] but we've lived with that since the inception of tramp/rcp/rssh and I don't consider it a problem personally. It's hard to support, requires hacking the innards of other packages and introduces load-order dependencies in the packages. Correct me if I'm wrong, but it's only a problem with EFS. You are wrong. It's a problem with Ange-FTP and EFS equally. EFS seems to be a little more willing to dig it's claws in about the issue, on occasion, but they /both/ suffer the same issue. Specifically, the FTP file access package assumes that `/[^/]:' paths belong to them. This means that we need to chose from: 1. Change away from using paths that match that. 2. Force EFS or Ange-FTP to be loaded before TRAMP is. I tend to talk about EFS more than Ange-FTP because I am an XEmacs user, so it's the tool I fight with more often. It's not because Ange-FTP doesn't suffer the same problems. However, I agree that we should fix it, if we can. Yes. Daniel Footnotes: [1] Well, at least /popularly/ chosen. :) -- Men are afraid of virgins, but they have a cure for their own fear and the virgin's virginity: fucking. Men are afraid of crones, so afraid of them that their cure for virginity fails them; they know it won't work. -- Ursula K. Le Guin
Re: TRAMP2 path definitions.
On 14 Mar 2001, Kai Grojohann wrote: On Wed, 14 Mar 2001, Edward J. Sabol wrote: On 15-Mar-2001, Daniel Pittman wrote: [...] It's hard to support, requires hacking the innards of other packages and introduces load-order dependencies in the packages. Correct me if I'm wrong, but it's only a problem with EFS. However, I agree that we should fix it, if we can. I think some packages look for `/foo:' at the beginning of a file name and then decide it's remote. I think. Hm. Maybe VC does this. Even in Emacs. Yes. Despite `file-remote-p'[1] being available, many packages are hard-coded to the knowledge of FTP-style paths. No matter what the Tramp filename format, I think these packages need to be changed, because (a) knowing that a file is remote does not necessarily tell you what you can do with it -- ange-ftp and Tramp offer different file operations.; and Yes. Especially because, for example, VC will refuse to operate on a FTP file path (because it can't), while TRAMP supports this. [...] But of course EFS is really really really pushy. It hooks itself into dired in many places (it has its own dired, in fact). Quite annoying. But I have spoken to the True. I intend to take this up with the maintainers of it, eventually. It's also true, though, of Ange-FTP. Daniel Footnotes: [1] Which /isn't/ a file operation, dammit, but a hard-coded function that *knows* about EFS or Ange-FTP.[2] [2] My hack makes this a file operation (yay), in the hope that I can convince the XEmacs beta team about it... and that someone can let me know how to get the same thing done with Emacs. -- Reality is a collective hunch. -- Jane Wagner
