summaryrefslogtreecommitdiff
path: root/support/vvcode/doc/vvintro.tex
blob: 303a08ce480aa3768f03599dfe6df595540d7712 (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
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
@c *************************************************************************
@c CHAPTER: Introduction to VVcode
@c *************************************************************************
@c    node-name, next, previous,  up
@node Introduction, Introduction, Top, Top
@chapter Introduction to VVcode

Reliable and faithful exchange of binary files between computers over
networks is a well-known problem, especially if the computers use different
operating systems and are connected to different networks via a gateway.
Unfortunately inter-networking and electronic mail are very much children
of the 1960s: they might have had to wait until the 1970s for their
naissance, but their progenitors were mentally locked-in to the concept of
the 7-bit ASCII code for conveying textual information.  The @TeX{}
community has long been aware of this problem when trying to exchange
``machine-independent'' @file{.dvi} files and font-related data such as
@file{.tfm} and @file{.pk} files.  It has sometimes been possible to
exchange this binary data by using encoding schemes that allow the data to
be represented using a subset of the seven-bit ASCII character set.

Academics and authors in many fields have hitherto been able to pass
@file{.tex} files back-and-forth by electronic mail---apart from a few
minor quirks and blemishes, such @TeX{} source files pass unharmed across
the planet's networks.  Problems are encountered when mail passes through
certain gateway machines which introduce irreversible character
corruptions.  Particularly notorious is the Janet/Bitnet gateway
which has the unfortunate habit of converting @samp{^} to @samp{~} and
@samp{~} to @samp{%}: since it leaves @samp{%} itself unaffected, this
makes recovery of the original file a non-trivial exercise. It sometimes
also changes the brace characters @samp{@{@}} into odd characters above
128: this is particularly embarrassing, of course, for @file{.tex} files!

For some years many @TeX{} users, particularly those working in languages
other than English, and thus familiar with character set encodings
containing other than the basic ASCII set, have been agitating for
@TeX{} to be able to handle input in their mother tongues, using their own
languages' character sets.  In 1989, Knuth announced @TeX{} V3, and
implementors world-wide beavered away to bring each implementation
up-to-date.  @TeX{} V3 now supports eight-bit character sets and so
@file{.tex} source files are now effectively `binary' files and will
therefore suffer from the same exchange problems experienced with
@file{.dvi} files.

All those authors that had previously been able to cooperate, despite being
separated by hundreds or thousands of miles, might once again be forced to
entrust floppy disks to the vagaries of the world's postal systems
(although one shouldn't underestimate the bandwidth of the Royal [or other]
Mail system).

Unless or until the various e-mail protocols, networks and software are
converted to support uncorrupted transmission of characters codes
@code{0x20 @dots{} 0x7e} and @code{0xa1 @dots{} 0xfe}, it will have to
become the norm for @file{.tex} sources to be encoded for transmission by
e-mail.

This problem is of course well known outside the @TeX{} community.


@c =========================================================================
@c SECTION: The Aston Archive
@c =========================================================================
@section{The Aston Archive}
The author is a volunteer assistant to Peter Abbott in running the world's
principal repository of @TeX{}-related material at Aston University in
Birmingham.  The archive (host: @code{TeX.Ac.Uk}) holds several hundred
megabytes of text and binary files including:

@itemize @bullet
  @item program sources for @TeX{}, @code{METAFONT}, DVI drivers and many
        other utilities;
  @item binary executables for a variety of popular operating systems (e.g. 
        Atari, Macintosh, MS-DOS, Unix, VAX/VMS and VM/CMS);
  @item @code{METAFONT} sources for Computer Modern and other fonts;
  @item binary font files (mainly @file{.tfm} and @file{.pk}) for a number of
	different output devices;
  @item text macro and style files.
@end itemize

The archive provides access to these files via the following services:

@itemize @bullet
  @item NIFTP@footnote{Network Independent File Transfer Protocol --- in the
                       UK, one does not perform the pseudo-login that
                       Internet users are accustomed to using with the FTP
                       protocol: instead, one issues a ``transfer request''
                       for a file to be sent to or from the remote
                       machine --- the transfer itself takes place
                       asynchronously. One nice consequence is that such
                       transfers can be queued for overnight execution,
                       leaving daytime bandwidth free for e-mail and true
                       remote interactive logins.}
        from Janet hosts---typically 300 megabytes of data are transferred
	every month; this would probably be much greater if we were not
	limited by the bandwidth of our 9600Bd connection to Janet.
  @item FTP and Telnet access from Internet hosts.
  @item Interactive browsing service via Janet PAD, including the facility
        to send files out using NIFTP (and later FTP).
  @item Interactive browsing service via dialup modem lines, including the
        facility to download files using Kermit and similar protocols.
  @item An e-mail file server which typically sends 150 megabytes of data per
        month to sites all over the world (though predominantly to
        EARN/Bitnet sites).
  @item A magnetic media distribution service via surface carriers.
        Copies of the entire archive have been sent to embryonic @TeX{}
        communities in Czechoslovakia, Hungary and Poland.
@end itemize

We have experienced many problems trying to support all of these file
types, operating systems and access methods.  The e-mail file server
clearly needs a reliable method of encoding files if its many customers are
not to be denied access to the non-text files in the archive.

Binary files such as @file{.pk} font files are stored in different ways to
accommodate the requirements of the different operating systems supported.
Currently we maintain multiple font directory trees for the Macintosh,
MS-DOS, Unix and VAX/VMS with all the attendant problems of
synchronization, disk space and archivists' time.  We need a single storage
format which allows export to all of our supported operating systems.


@c =========================================================================
@c SECTION: Specification for a Coding Scheme
@c =========================================================================
@section{Specification for a Coding Scheme}
In mid-1990, the archivists came to the conclusion that a universal
encoding scheme was required to accommodate the many different kinds of
file and file organizations that needed to be supported by the archive.

Niel Kempson formulated the first draft of this specification in mid-1990;
the requirements of the encoding scheme may be summarized as follows:

@table @strong
  @item Preserving File Structure
        It is insufficient, especially for an archive holding text and
        binary files for a variety of machine types, merely to encode data
        simply as a stream of bytes:

      @itemize @bullet
	@item Virtually all operating systems (except Unix) make a
              distinction between binary and text files, so the coding
              system should recognize and maintain this distinction.
	@item Unix and most PC-based operating systems treat files as
              streams of bytes with no further structure imposed.  On the
              other hand, certain widely-used operating systems (e.g.
              VAX/VMS and VM/CMS) have record-oriented file systems where
              different types of file are stored in a format appropriate to
              the type of file@footnote{It is often argued that the
              increase in efficiency more than offsets the increase in
              complexity.}.

              For these operating systems, we consider it essential that
              the encoding scheme should identify, preserve and record the
              most commonly used file organizations.   The decoding program
              should be able to use this information to create the output
              file using the organization appropriate to the operating
              system in use.  If the information is of no consequence to
              the receiving system, the default file structure (if any)
              should be created.  If the encoding system does not have
              structure in its files, the receiving system may provide
              suitable defaults automatically.  In all cases the programs
              should permit the user to override or supplement file
              structure information.

	@item Whenever possible, these details of structure should be
              determined automatically by the encoding program; at the very
              least, an indication of whether the file is text or binary
              shall be provided, even under an operating system such as
              Unix that need make no such distinction for its own use, to
              allow decoding to an appropriate file organization on those
              systems that {@emph do} make such a distinction.
      @end itemize

  @item Coding Scheme
        Whatever method is used must allow encoded data to be e-mailed:

      @itemize @bullet
	@item It should be possible to specify the coding table to be used
              to encode the data.  The coding table used shall be recorded
              with each part of the encoded data.
	@item If a recorded coding table is found while decoding, it should
              be used to construct an appropriate decoding table.  Simple
              one-to-one character corruptions should be corrected as long
              as only one of the input characters is mapped to any one
              output character.
	@item The recommended encoding uses only the following characters:
              @quotation
	        @code{+-0123456789}@*
	        @code{abcdefghijklmnopqrstuvwxyz}@*
	        @code{ABCDEFGHIJKLMNOPQRSTUVWXYZ}
              @end quotation

	      Such an encoding as originally used for XXcode has been shown
              to pass successfully through all the gateways which are known
              to corrupt characters.
      @end itemize

  @item Integrity of Encoded Data
        We want to ensure that the @emph{whole} encoded file passes through
        the e-mail network.

      @itemize @bullet
	@item Encoded lines should be prefixed by an appropriate character
              string to distinguish them from unwanted lines such as mail
              headers and trailers.  Whilst not essential, this feature
              does assist the decoding program in ignoring these spurious
              data.
	@item Lines should not end with whitespace characters as some
              mailers and operating systems strip off trailing whitespace.
	@item The encoding program should calculate parameters of the input
              file such as the number of bytes and CRC and record them at
              the end of the encoded data.
      
	      The decoding program should calculate the same parameters
              from the decoded data and compare the values obtained from
              those recorded at the end of the encoded data.
      @end itemize
   @item Making Files Mailable
         A mechanism is needed to overcome some gateways' refusal to handle
         large files.

      @itemize @bullet
	@item The encoding program should be able to split the encoded
              output into parts, each no larger than a maximum specified
              size.  Splitting the output into smaller parts is useful if
              the encoded data is to be transmitted using electronic mail
              or over unreliable network links that do not stay up long
              enough to transmit a large file.  The recommended default
              maximum part size is 30kB.
	@item The decoding program should be able to decode a multi-part
              encoded file very flexibly.  It should @emph{not} be
              necessary to:

	    @enumerate
		  @item strip out mail headers and trailers;
		  @item combine all of the parts into one file in the
			correct order;
		  @item process each part of the encoded data as a
			separate file.
	   @end enumerate
        @item In addition any file specifications from the operating system
              on which the VVE file was created must not prevent the file
              from being decoded.
      @end itemize

  @item Miscellaneous
        Further considerations include:

    @itemize @bullet
      @item Support for character sets other than ASCII is essential
            if the encoding scheme is to be useful to IBM hosts.  The
            encoding program should label the character set used by the
            encoded data, and both encoder and decoder should enable the
            conversion between the local character set and another
            character set.  For example a user on an EBCDIC host
            should be able to encode text files for transmission to another
            EBCDIC host, or to convert them to ASCII before
            encoding and transmission to an ASCII host. Similarly,
            that user should be able to decode text files from ASCII
            and EBCDIC machines, creating EBCDIC output files.
      @item Where possible, the original file's timestamp should be encoded
            and used by the decoding program when recreating the file: this
            will permit archives to retain the originator's time of
            creation for files, and thus permit the users (not to mention
            the archivists) to identify more clearly when a new version of
            a file has been made available.  Timezones should be supported
            where possible.
      @item The encoding and decoding schemes should be able to read and
            write files that are compatible with one or more of the well
            established coding schemes (e.g. UUcode, XXcode).
      @item The source code for the programs should be freely available.
            It should also be portable and usable with as many computers,
            operating systems and compilers as possible.
    @end itemize
@end table


@c =========================================================================
@c SECTION: The Search Commences
@c =========================================================================
@section{The Search Commences}
Naturally, the first step was to examine the existing coding schemes in
comparison with the above ideal specification.  Such schemes fell into two
broad classes: @dfn{portable schemes}, which were intended to permit the
encoding of files on any computer architecture into a form that could be
transmitted electronically, and decoded on the same or a different
architecture; and @dfn{platform-specific schemes}, which provided rather
better support for transferring files between two computers using the same
architecture and operating system.

@subsection{Portable Coding Schemes}
The most commonly used coding schemes supported by a variety of platforms
are:

@itemize @bullet
  @item @code{BOO}
  @item @code{UU}
  @item @code{XX}
@end itemize

Most implementations of these schemes known to the authors are designed for
use with stream file systems.  These programs have no means of recording,
let alone preserving, record structure and are thus unsuitable for our
purposes.  This is not surprising since @code{UUcode} and its mutation
@code{XXcode} were developed specifically for exchanging files between Unix
systems.  In fairness to these schemes, they are well suited to the
transmission of text files and certain unstructured binary files.  

Standard @code{UUcode} encodes files using characters @samp{ } @dots
@samp{_} of ASCII.  This can result in one or more spaces appearing at
the ends of lines: some mailers decide that this is information not worth
transmitting, with consequent inability to reconstruct the original file.

Files containing characters such as @samp{^} are often irreversibly
corrupted by mail gateways; this problem led to the development of
@code{XXcode} which uses a rather more robust character set, namely:

@quotation
  @code{+-0123456789}@*
  @code{abcdefghijklmnopqrstuvwxyz}@*
  @code{ABCDEFGHIJKLMNOPQRSTUVWXYZ}
@end quotation

The encoding table used is recorded with the encoded data to allow the
detection of character corruptions, and the correction of reversible
character transpositions.  Whilst superficially a step forward,
@code{XXcode} offered little more than most existing versions of
@code{UUcode}, which already supported coding tables.  Its major
contribution was in formalizing the encoding table, and in particular its
default table was proof against all the known gateway-induced corruptions.


@subsection{Platform Specific Coding Schemes}

Encoding schemes have been developed to support transfer of files
possessing some structure which therefore cannot be reconstructed correctly
when encoded by the portable schemes.  When the encoding and decoding
programs of such a platform specific scheme are each used on the same
computer and operating system type, files may be encoded and transmitted
with a great deal of confidence that the decoded file will reproduce the
original's structure and attributes in their entirety.

Examples of such programs are @code{TELCODE} and @code{MFTU}  for VMS,
@code{NETDATA} for IBM mainframes, and @code{Stuffit} and @code{MacBinary}
for the Macintosh.  But these programs have the major disadvantage that
they have each been implemented @emph{only} on the single architecture for
which they were designed: thus the only two of these schemes that could be
used on the VMS-based Aston Archive would be of minimal interest elsewhere!

The Archive's content is in some respects artificially inflated by the
presence of @file{.hqx} files for Macintoshes, @file{.boo} for MS-DOS,
etc., which have to be held in pre-encoded form for transfer by those
requiring them.


@c =========================================================================
@c SECTION: VVcode is Born
@c =========================================================================
@section{VVcode is Born}

Realizing that none of the existing portable schemes were close enough to
our ideal, an early version of our specification was circulated on various
mailing lists by Niel Kempson towards the end of 1990. When the anticipated
``nil return'' was all that resulted, Brian Hamilton Kelly went ahead and
created a rudimentary @code{VVencode} by modifying an existing VAX Pascal
implementation of @code{uuencode}. After generating the companion
@code{VVdecode}, he then re-implemented the programs in Turbo C under the
MS-DOS operating system on the IBM-PC, and thereby was able to prove that
the new scheme was both viable and sufficient.

This version didn't support file formats, time stamping, file splitting,
character sets or CRC checking.


@subsection{A Production VVcode}

Following the minor feasibility study, Niel Kempson re-engineered the pair
of programs from scratch (adding certain features of the evolving
specification), paying particular attention to making the code portable
across a wide variety of operating systems.  Particular care was taken to
avoid the use of supposedly ``standard'' C functions that experience had
shown behaved differently under individual manufacturer's implementations,
or were even non-existent in some.  Therefore the code may sometimes appear
to be performing certain operations in a very long-winded way; it's very
easy to look at it and say ``why didn't the author use the @code{foo()}
function, which does this much more efficiently?'', but this function may
not even exist under another implementation of C, or behave in a subtly
different manner.

The core functions of @code{VVcode} are implemented as a collection of
routines written in as portable a fashion as possible, and a separate
module of a few operating system specific routines for file I/O,
timestamping, command-line or other interface, etc.  Porting @code{VVcode}
to a new platform should require only that this latter module be
re-implemented, in most cases by adapting an existing one.  

@code{VVcode} implements all of the features listed in the specification,
apart from the ability to generate @code{UUcode} and @code{XXcode}
compatible files.  However, the decoding program is backwards compatible
and can decode files generated by @code{UUcode} and @code{XXcode}.


@subsection{Arguments against VVcode}

When the advent of the @code{VVcode} system was first aired in the various
electronic digests, some heated debate followed along the lines that a new
encoding scheme was unnecessary, since @code{UUcode}/@code{XXcode} sufficed
@emph{for them}.  However, all these correspondents were Unix users who had
interpreted the @samp{VV} as meaning ``Vax-to-Vax'' by analogy with
@samp{uu}@footnote{@samp{V} was chosen simply because it followed @samp{U};
at one time, we'd seriously considered calling it @code{YAFES} --- Yet
Another File Encoding Scheme!} and who felt that such a scheme should be
private to VAXen.  The authors' reply was to the effect that the encoding
scheme was intended to support the needs of archives like Aston's, and as
such, had to provide

@enumerate
    @item an automated tool (it would be somewhat difficult to expect our
          users to be able to tell the encoder what sort of file structure
          it was handling, when this concept was entirely alien to many of
          them);
    @item facilities to encode binaries for many operating systems;
    @item mail server features, such as splitting of large files;
    @item operation across the widest possible combination of platforms.
@end enumerate

The overhead of using the @code{VVcode} system is at most a couple of
hundred bytes over using @code{UUcode}, and the extra functionality and
@emph{universality} with respect to @code{UUcode} or @code{XXcode} thereby
comes almost for free.  


@c =========================================================================
@c SECTION: 
@c =========================================================================
@section{Availability of VVcode}

At present, the @code{VVcode} system is only available in C, but it has
been shown to run successfully on the following combinations of hardware,
operating system and compiler:

@table @strong
  @item Macintosh
        At the time of writing (May 1991) John Rawnsley of the University
        of Warwick had commenced development of a Macintosh port, which
        will encode the resource and data forks in a manner that will
        permit the former to be ignored by non-Macintosh systems.

  @item MS-DOS
    @itemize @bullet
      @item IBM PS/2, PC (and clones); MS-DOS 3.3, 4.01, 5.00;
            Borland Turbo C 1.5, 2.0, Borland C++ 1.0, 2.0, 3.0 and
            Microsoft C 5.1, 6.0
    @end itemize

  @item OS/2
    @itemize @bullet
      @item IBM PS/2, PC (and clones); OS/2 2.0; Microsoft C 6.0 and 
            GNU C 2.1
    @end itemize

  @item Unix
    @itemize @bullet
      @item Sun 3; SunOS 3.x and 4.0.3; native C and GNU C
      @item Sun Sparcstation 1; SunOS 4.1; native C and GNU C
      @item SCO Unix V/386 v3.2.2, Microsoft C compiler
    @end itemize

  @item VAX/VMS
    @itemize @bullet
      @item All VAXen; VMS 5.2--5.4-1; VAX C V3.0--V3.2 and GNU C 1.40
    @end itemize

  @item VM/CMS
    @itemize @bullet
      @item VM/CMS; Whitesmith C compiler v1.0 (This implementation was
            ported by Rainer Sch@"opf; basing it upon the Unix
            implementation, this took him about one day.)
    @end itemize

@end table