summaryrefslogtreecommitdiffstats
path: root/kioslave/fish/README
blob: d1afdc3d1f5133db29597d6d8b3af1b18012c10b (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
Overview of kio_fish
====================

  ------------------------------------------------------------------------
  NOTE FOR KDE2 USERS: This is the last release supporting KDE2. However,
  you might need to modify Makefiles to get things installed into the
  right directories.
  ------------------------------------------------------------------------

  FISH is a protocol to get filesystem access without special server
  software, only using a remote shell. (Hence the name: FIles transferred
  over SHell protocol).
  It was first devised by Pavel Machek <pavel@bug.ucw.cz> and implemented
  as a Midnight Commander vfs module in 1998.

  This is a complete client implementation using his original version
  0.0.2 protocol, extending it with 2 commands (which are only optional -
  should a real FISH server exist on server side that doesn't understand
  them, this ioslave still works, only slower). Moreover, this client does
  complete shell metacharacter quoting on all arguments, a fact that is
  neccessary but missing from the specs.
  Extensions used are: append (APPEND command), copy (COPY command),
  lscount (LIST first prints number of files to be listed), lslinks (LIST
  shows symlink info instead of info about link targets), lsmime (LIST
  determines the MIME type on the server side)
  Password and host key queries are handled via dialog boxes.
  The goal of this client is to make a remote directory look and feel exactly
  like a local directory, with all comfort, only slower.

  NOTE: From version 1.1.3 on, compression is no longer turned on auto-
  matically. You have to specify it via ~/.ssh/config or wherever
  your local ssh client reads its settings. The same goes for all other
  connection parameters. OpenSSH for example has a powerful configuration
  file syntax which lets you configure access differently for each host,
  something I do not intend to duplicate. Read the ssh_config(5) man page
  for details. If someone knows the docs to read for commercial ssh please
  tell me so I can include that here as well.

  Included below is the original posting from the mc mailing list archives.
  
  If perl is installed on the remote machine and in the default PATH, it will
  be used to transfer a custom server script which is much faster than
  shell-only mode and more predictable as well. The script is stored in a
  file called .fishsrv.pl in the working directory directly after login and
  will be reused on subsequent connections.

  2001/10/07 Jörg Walter <trouble@garni.ch>



From: Pavel Machek <pavel@bug.ucw.cz>
Subject: New virtual filesystem - fish
Date: Tue, 15 Sep 1998 22:30:07 +0200

Hi!

New virtual filesystem has been created, which allows you to access
files on remote computer over rsh/ssh connection, with _no_ server
needed on the other side. To use it from mc or any program using
libvfs.so, do

cd /#sh:user@host.to.connect.to/

Note that password authentication will not work, so you must be
authenticated using [rs]hosts or RSA key.

For protocol, see mc/vfs/README.fish. If someone wants to write
server, it would be good idea, since it works without server but
performance is not optimal.

							Pavel

PS: Protocol looks like this. If you have any comments, it is time to
speak.


		FIles transferred over SHell protocol (V 0.0.2)
		~~~~~~~~~~~~~~~~~~~~~~~~~~~~

This protocol was designed for transferring files over secureshell
(ssh) connection. It can be as well used for transfers over rsh, and
there may be other uses.

Client sends requests of following form:

#FISH_COMMAND
equivalent shell commands,
which may be multiline

Only fish commands are defined here, shell equivalents are for your
information only and will probably vary from implementation to
implementation. Fish commands always have priority: server is
expected to execute fish command if it understands it. If it does not,
however, it can try the luck and execute shell command.

Server's reply is multiline, but alwyas ends with

### 000<optional text>

line. ### is prefix to mark this line, 000 is return code. Return
codes are superset to those used in ftp.

There are few new exit codes defined:

000 don't know; if there were no previous lines, this marks COMPLETE
success, if they were, it marks failure.

001 don't know; if there were no previous lines, this marks
PRELIMinary success, if they were, it marks failure

				Connecting
				~~~~~~~~~~
Client uses "echo FISH:;/bin/sh" as command executed on remote
machine. This should make it possible for server to distinguish FISH
connections from normal rsh/ssh.

				Commands
				~~~~~~~~
#FISH
echo; start_fish_server; echo '### 200'

This command is sent at the begining. It marks that client wishes to
talk via FISH protocol. #VER command must follow. If server
understands FISH protocol, it has option to put FISH server somewhere
on system path and name it start_fish_server.

#VER 0.0.2 <feature1> <feature2> <...>
echo '### 000'

This command is the second one. It sends client version and extensions
to the server. Server should reply with protocol version to be used,
and list of extensions accepted.

VER 0.0.0 <feature2>
### 200

#PWD
pwd; echo '### 200'

Server should reply with current directory (in form /abc/def/ghi)
followed by line indicating success.

#LIST /directory
ls -lLa $1 | grep '^[^cbt]' | ( while read p x u g s m d y n; do echo "P$p $u.$g
S$s
d$m $d $y
:$n
"; done )
ls -lLa $1 | grep '^[cb]' | ( while read p x u g a i m d y n; do echo "P$p $u.$g
E$a$i
dD$m $d $y
:$n
"; done )
echo '### 200'

This allows client to list directory or get status information about
single file. Output is in following form (any line except :<filename>
may be ommited):

P<unix permissions> <owner>.<group>
S<size>
d<3-letters month name> <day> <year or HH:MM>
D<year> <month> <day> <hour> <minute> <second>[.1234]
E<major-of-device>,<minor>
:<filename>
L<filename symlink points to>
<blank line to separate items>

Unix permissions are of form X--------- where X is type of
file. Currently, '-' means regular file, 'd' means directory, 'c', 'b'
means character and block device, 'l' means symbolic link, 'p' means
FIFO and 's' means socket.

'd' has three fields: month (one of strings Jan Feb Mar Apr May Jun
Jul Aug Sep Oct Nov Dec), day of month, and third is either single
number indicating year, or HH:MM field (assume current year in such
case). As you've probably noticed, this is pretty broken; it is for
compatibility with ls listing.

#RETR /some/name
ls -l /some/name | ( read a b c d x e; echo $x ); echo '### 100'; cat /some/name; echo '### 200'

Server sends line with filesize on it, followed by line with ### 100
indicating partial success, then it sends binary data (exactly
filesize bytes) and follows them with (with no preceeding newline) ###
200.

Note that there's no way to abort running RETR command - except
closing the connection.

#STOR <size> /file/name
<i><font color="#008000">> /file/name; echo '### 001'; ( dd bs=4096 count=<size/4096>; dd bs=<size%4096> count=1 ) 2>/dev/null | ( cat > %s; cat > /dev/null ); echo '### 200'
</font></i>
This command is for storing /file/name, which is exactly size bytes
big. You probably think I went crazy. Well, I did not: that strange
cat > /dev/null has purpose to discard any extra data which was not
written to disk (due to for example out of space condition).

[Why? Imagine uploading file with "rm -rf /" line in it.]

#CWD /somewhere
cd /somewhere; echo '### 000'

It is specified here, but I'm not sure how wise idea is to use this
one: it breaks stateless-ness of the protocol.

Following commands should be rather self-explanatory:

#CHMOD 1234 file
chmod 1234 file; echo '### 000'

#DELE /some/path
rm -f /some/path; echo '### 000'

#MKD /some/path
mkdir /some/path; echo '### 000'

#RMD /some/path
rmdir /some/path; echo '### 000'

#RENAME /path/a /path/b
mv /path/a /path/b; echo '### 000'

#LINK /path/a /path/b
ln /path/a /path/b; echo '### 000'

#SYMLINK /path/a /path/b
ln -s /path/a /path/b; echo '### 000'

#CHOWN user /file/name
chown user /file/name; echo '### 000'

#CHGRP group /file/name
chgrp group /file/name; echo '### 000'

#READ <offset> <size> /path/and/filename
cat /path/and/filename | ( dd bs=4096 count=<offset/4096> > /dev/null;
dd bs=<offset%4096> count=1 > /dev/null;
dd bs=4096 count=<offset/4096>;
dd bs=<offset%4096> count=1; )

Returns ### 200 on successfull exit, ### 291 on successfull exit when
reading ended at eof, ### 292 on successfull exit when reading did not
end at eof.

#WRITE <offset> <size> /path/and/filename

Hmm, shall we define these ones if we know our client is not going to
use them?


That's all, folks!
						pavel@ucw.cz


--
I'm really pavel@atrey.karlin.mff.cuni.cz. 	   Pavel
Look at http://atrey.karlin.mff.cuni.cz/~pavel/ ;-).