Commit | Line | Data |
---|---|---|
9eb668c9 BJ |
1 | Network Status January 11, 1980 |
2 | ||
3 | BUGS | |
4 | ---- | |
5 | -- Various response messages are lost. This includes "fetching" | |
6 | files when the file being retrieved never arrives. I suspect | |
7 | this has something to do with unreliable delivery of error messages, | |
8 | but this is not reliably reproducible. | |
9 | ||
10 | -- The net command will create files in the queue directories | |
11 | without the corresponding control files ("dfa..." without "cfa..."). | |
12 | Unknown cause. They should be periodically removed. | |
13 | (Perhaps caused by an error such as an invalid machine name.) | |
14 | ||
15 | -- The network makes no provision for errors in transit on intermediate | |
16 | machines, such as "No more processes" or "File System Overflow". | |
17 | While these occur only rarely, when they do, no message or | |
18 | notification is sent to anyone. | |
19 | ||
20 | -- The network rendezvous protocol seems to occasionally get | |
21 | in a state where a specific file is continually retransmitted | |
22 | and never seems to get through. This happens when both the | |
23 | host system and the network queues are overloaded, and thus | |
24 | is very unpleasant to debug. | |
25 | ||
26 | -- The network daemons occasionally core dump. They should not. | |
27 | ||
28 | ||
29 | SUGGESTIONS | |
30 | ----------- | |
31 | ||
32 | -- Performance Improvements: | |
33 | A number of links now running at 1200 Baud could have their speeds | |
34 | changed to 9600 Baud without deteriorating the system the | |
35 | network runs on. | |
36 | There are some high speed links (dmc-11's) which the network | |
37 | could use for much better performance. | |
38 | Likewise, the Bussiplexor could be used as a faster link. | |
39 | This would allow us to increase the present 100,000 character | |
40 | file length limit. | |
41 | All the links would be faster if UNIX kernel drivers were used to avoid | |
42 | going through the terminal character queues and interrupting | |
43 | the CPU for every character. | |
44 | At the end of every quarter, network transmission speed decreases and | |
45 | the volume of traffic increases. The network becomes saturated | |
46 | between two links and requests may arrive days later. | |
47 | Increases in the link speed would reduce these seasonal delays | |
48 | a great deal. | |
49 | ||
50 | -- Maintenance Improvements: | |
51 | The network has become large enough to make re-compilation | |
52 | of the source on all machines to become practically impossible. | |
53 | The net command has compiled within it a routing table for each | |
54 | remote machine (defined in config.h). | |
55 | Adding a new machine to the network requires recompiling the | |
56 | net command on ALL machines. The net command should read an | |
57 | external text file to compute its data structures. | |
58 | There is a program patchd, written by Bill Joy, which could | |
59 | be used to patch the binary versions of the network | |
60 | on like-systems, such as the Computer Center machines. | |
61 | The network code should use the retrofit library for | |
62 | non-Version 7 systems. | |
63 | ||
64 | -- Network mail needs to be generalized in a number of ways. | |
65 | People with accounts on many machines want their mail forwarded | |
66 | to one specific machine. Also, there are at least two other networks | |
67 | now connected to the Berkeley network (the Bell Research net and | |
68 | the Arpanet), and mail destined for those networks should be | |
69 | routed to the appropriate gateway. Neither of these is particularly | |
70 | difficult to implement, but system mail is an important facility | |
71 | and the people in charge of the various machines on the network | |
72 | disagree on how these features are to be added, especially concerning | |
73 | issues of reliability and error reporting. | |
74 | ||
75 | -- The possibility of a number of small UNIX personal machines wanting | |
76 | intermittent access to the network looms ahead. We should attempt | |
77 | to organize the software to allow occasional use | |
78 | by other UNIX machines, without tying down a port all the time. | |
79 | ||
80 | -- The A machine has a typesetter that can be used from the | |
81 | Computer Center machines through the network. It would be nice | |
82 | if this facility were available from non-Computer Center machines | |
83 | to the A machine. Programs exist to provide this and have been used | |
84 | extensively by Bill Joy and myself but the | |
85 | Computer Center is reluctant to open up that facility for | |
86 | security and reliability reasons. | |
87 | We would like to arrange for Computer Center job numbers to | |
88 | be stored in the password file on non-CC machines, to | |
89 | allow people without accounts on A to have access to the | |
90 | typesetter. | |
91 | ||
92 | -- Bob Fabry has suggested the "machine" be generalized to imply a | |
93 | machine/account pair, e.g. -m caf would imply "caf" on Cory, | |
94 | -m Cory would imply "fabry" on Cory. | |
95 | Environments could provide this information. | |
96 | It has also been suggested that the notion of a "default" machine | |
97 | is too restrictive and that each type of command should have a | |
98 | default machine, e.g. netlpr to A, net to B, netmail to C, etc. | |
99 | ||
100 | -- Colin has developed some data compression algorithms. On machines | |
101 | which are normally CPU idle, his algorithms could be used to | |
102 | compress data and speed up file transfer. | |
103 | Each individual host could decide whether data should be compressed, | |
104 | and each receiving machine would be able to handle both compressed | |
105 | and uncompressed data. | |
106 | ||
107 | -- Files being retrieved, or fetched, are created zero-length | |
108 | as the request is sent to the remote machine. An alternative | |
109 | would be to put the message "File being transferred." in the file to | |
110 | make things clearer. | |
111 | ||
112 | -- File modes should be preserved across the network. Currently | |
113 | they are set to 0600 most of the time. | |
114 | ||
115 | -- It would be nice if the rcs facilities and commands on various | |
116 | UNIX machines with rcs links were more accessible from machines | |
117 | without an rcs link. | |
118 | ||
119 | -- The network was not expected to become as large as it has. | |
120 | Not much thought was given to large networks. | |
121 | The netq command only lists queues on the local machine, | |
122 | but many times the user is waiting for long queues on intermediate | |
123 | machines. | |
124 | Likewise, once the request is forwarded to the nearest machine, | |
125 | the netrm command will not let the originator remove the queue file. | |
126 | Finally, a network status command telling people what the network | |
127 | was doing would be very helpful. | |
128 | ||
129 | -- The file length restriction of 100,000 characters forces users to split | |
130 | their files up into small pieces. The network should have a | |
131 | way to do this split automatically. | |
132 | ||
133 | -- The underlying protocol is wasteful and/or confusing in a | |
134 | number of ways: | |
135 | * The request length should be in ASCII, not a long integer. | |
136 | * Remove the extra 5 character string at the beginning of each | |
137 | transmission. | |
138 | * Compute a full checksum on the entire file in addition | |
139 | to the checksum per packet now provided. | |
140 | It is unlikely these will be changed since all the daemons | |
141 | on the network machines would have to be changed at once. | |
142 | ||
143 | -- The netcp command should allow the user to default one of | |
144 | the filenames to a directory, ala the cp command. | |
145 | ||
146 | -- File transfers, like remote mail, should be possible from | |
147 | the Berkeley Network to the Arpanet and the Bell Research Net. | |
148 | This is not difficult technically, but requires UNIX-like | |
149 | stream interfaces to be written for the gateways. | |
150 | ||
151 | -- Currently the network files being transferred are | |
152 | copied into /usr/spool... it would be nice for | |
153 | large files to simply use a pointer to them. | |
154 | (To save time and space). | |
155 | ||
156 | -- The scheduler the daemon uses is very simple. | |
157 | It should have a way to age priorities and to "nice" | |
158 | transfers, to be done after all normal ones are done. | |
159 | Also, there are some network uses that are time-dependent. | |
160 | It would be nice if certain queue files would disappear | |
161 | at certain times, if for example, a remote machine were down, | |
162 | given that they are no longer useful. |