dick@tjalk.UUCP (Dick Grune) (04/24/86)
1. INTRODUCTION At this moment we, the authors, have at our disposal 7 emacses, all slightly or more than slightly different in their key bindings, enough to confuse novice and expert alike. Emacs's great strength, its programmability, turns into a weakness and chaos rages. In an attempt to bring some order into this chaos we have drawn up the complete lists of initial key bindings of each emacs (i.e. the bindings the unsuspecting user gets when he just calls the specific emacs) and printed them side by side, on a virtual piece of paper, 210 columns wide. Upon examination, patterns began to emerge, and we have observed the following. 1. Almost all elementary functions, functions "that every emacs should have," are always bound to the same keys, ^H and ^Z being the only problems. These key bindings are brought together in Class A in the proposal. 2. Most of the non-essential functions have a "home", a binding about which a majority of the emacses agree. Those for which a home could be identified are grouped in Class B. 3. All emacses agree that ^X- or ESC- followed by a capital letter is either bound to nothing or to the same function as ^X- or ESC- followed by the corresponding small letter. 4. None of the emacses recognizes the problems caused by the fact that on some systems ^S and ^Q are preempted by the terminal communication, as X-OFF and X-ON. Any emacs should supply reasonable alternatives. This has led us to submit the attached proposal, which is accompanied by a rationale. We are very interested to receive reactions from any interested party. We will summarize to the net, or, if too many people think this is a bad idea, shut up. Dick Grune dick@vu44.UUCP Sjoerd Mullender sjoerd@vu44.UUCP Vrije Universiteit de Boelelaan 1081 1081 HV Amsterdam the Netherlands 2. PROPOSAL Proposal for recommended key bindings in emacs-like editors. SCOPE The recommendations in this proposal apply to the key bindings as they appear initially --upon calling the editor-- to the user who has given no customization or initialization information. This proposal is NOT to be construed as a recommendation against differing local or personal key bindings, and is intended solely to reduce confusion in the unsuspecting user and as matter for consideration to future emacs authors. The editor functions are grouped into 3 classes, according to their relative importance: Class A: functions the user can expect to be present. Class B: optional functions, which the user can nevertheless expect to be present in a wordprocessor/development environment. Class C: highly optional or specialized features, which are common enough for a recommendation to be in order. In addition to these classes, two general recommendations are given. Class A: Character commands: ^B backward character ^D delete next character ^F forward character ^T transpose characters ^? delete previous character Line commands: ^A beginning of line ^E end of line ^K kill to end of line Intra-line commands: ^N next line ^P previous line Word commands: ESC-b backward word ESC-d delete next word ESC-f forward word ESC-^? delete previous word Window commands: ^V next page ^Z scroll forward ^X-1 delete other windows ^X-2 split window ^X-d delete window ^X-n next window ^X-o other window ^X-p previous window ESC-^V page next window ESC-, beginning of window ESC-. end of window ESC-< beginning of file ESC-> end of file ESC-v previous page ESC-z scroll backward Buffer commands: ^X-^B list buffers ^X-b switch to buffer ^X-k kill buffer File commands: ^X-^F find file ^X-^I insert file ^X-^M write modified files ^X-^R read file ^X-^S save file ^X-^W write named file Region commands: ^@ set mark ^W delete region to internal buffer ^Y yank from internal buffer ^X-^X exchange cursor and mark White space commands: ^I tab ^J insert newline and indent ^M insert newline ^O insert newline and backup Search commands: ^R search reverse ^S search forward ESC-q query replace string ESC-r replace string Miscellaneous commands: ^C exit ^X-^C exit ^L redraw screen ^Q quote next character ^U argument prefix ^^ replacement for ^Q ^\ replacement for ^S (or vice versa?) ^] alternate prefix ^G abort ^X-^G abort ESC-^G abort Class B: Word commands: ESC-t transpose words Sentence commands: ^X-^? kill to beginning of sentence ESC-a backward sentence ESC-e forward sentence ESC-k kill sentence Paragraph commands: ESC-[ backward paragraph ESC-] forward paragraph ESC-j justify paragraph Window commands: ^X-^ enlarge window Region commands: ESC-w copy region to internal buffer ESC-^W delete region to end of internal buffer Case commands: ^X-^L lower case region ^X-^U upper case region ESC-c capitalize word ESC-l lower case word ESC-u upper case word White space commands: ^X-^O delete blank lines ESC-\ delete horizontal space Macro commands: ^X-( start keyboard macro ^X-) stop keyboard macro ^X-e execute keyboard macro Internal editor commands: ESC-ESC evaluate internal expression ESC-x execute internal command Information commands: ^X-= display cursor position ESC-? help System commands: ^X-^E compile ^X-^N next error ^X-! give command to command interpreter ^_ call command interpreter Class C: Line commands: ^X-^T transpose lines ESC-m go to first non-blank Buffer commands: ESC-~ buffer not modified Information commands: ^X-* print version number Abbreviations: ^X-^A add local abbreviation ^X-^H inverse add local abbreviation ^X-+ add global abbreviation ^X-- inverse add global abbreviation Layout control: ^X-. set indent column ^X-; set comment column ^X-f set fill column ESC-; indent to comment column Scroll sideways: ^X-< scroll left ^X-> scroll right Parentheses: ESC-^B backward parenthesis ESC-^F forward parenthesis CAPITAL LETTERS IN A PREFIXED COMMAND Key strokes consisting of a prefix (^X or ESC) followed by a capital letter should not be bound. If this is found very inconvenient, they should be bound to the same command as the prefix followed by the corresponding small letter. ^S AND ^Q All key bindings that include a ^S or ^Q should have a corresponding binding with ^\ substituted for ^S and ^^ for ^Q. If possible, this substitution should be hard-wired (to be switched off on request); if not, appropriate explicit bindings should be supplied. END OF PROPOSAL 3. RATIONALE 3.1 The concept The reason for this proposal at all is the same as that for any standard: some design decisions are to a large extend arbitrary, and to reduce the ensuing complexity, additional artificial order has to be imposed. We feel that the subject has at present reached a certain maturity and has not yet petrified. 3.2 The approach We have considered the initial key bindings of emacs2.10 (Unipress/Gosling), jove (Jonathan Payne/Unison World), PerfectWriter (Perfect Software), MicroEmacs (Curt Jutzi, jutz@pogo), GNU (Richard M. Stallman), scame (Leif Samuelsson, leif@erisun), MicroEMACS (conroy@dec-rex). For Class A and B we only considered those key bindings that were defined in at least four of the above. The key bindings for which no rationale is given below were either unanimous or defined by some and not contradicted by any of the others. We have not concerned ourselves with small semantic differences: "next-paragraph" will go to the beginning of the next paragraph on some editors and to the end of this paragraph on others; "transpose-characters" may transpose cursor[-2..-1] or cursor[-1..0], etc. These are considered the same, since the user has direct optical feedback. 3.3 Specific key bindings ^H is left unbound. It is usually bound to backwards-character or to delete-previous-character, without clear preference for either. Both of these are available under other keys (^B and ^?). Users working on low-grade telephone lines may need to avoid ^? since it is often generated by noise; in this case they may bind delete-previous-character to ^H and unbind ^?. ^Z Some systems have the possibility to suspend a running process; this action is traditionally coupled to the ^Z key. It is, however, not at all universal and requires considerable OS support. We feel therefore that we can still use ^Z (and its compagnion ESC-z) for scrolling --with suspending, calling command interpreters, etc., bound to ^_ . ^] Most editors do not use a third prefix besides ^X and ESC; among those that do, there is no consensus, with ^H, ^Z, ^\, ^^ and ^X-4 being used. ^] is not used by any editor for anything. ^X-^F, ^X-^R Generally there exist two commands, "read file into new buffer" and "read file into existing buffer", called "find file" and "read file" respectively here. They are bound to various permutations of ^X-^F, ^X-^R and ^X-^V, the latter being called "visit file", but being identical to "find file". The proposed assignment seems more mnemonic and seems to have a slight majority. 3.4 Capital letters in a prefixed command The beginning user would have difficulty in telling apart ^X-^S, ^X-s, ^X-S, ESC-^S, ESC-s and ESC-S, with assorted meanings like save-file, forward-search, call-spelling-checker, center-line or scratch-buffer. Some tutorials print ^X-S when they mean ^X-s. Some terminals with function keys send ESC-@ to ESC-O for them, which should remain free for special applications. 3.5 ^S and ^Q ^S and ^Q are used by some communication protocols and will then cause a great deal of confusion, both to the user and to the computer. It seems preferable to lift this problem out of the user realm and into the editor. It is disturbing that we have seen editors that have the mapping ^S -> ^\ ^Q -> ^^ and those that have ^S -> ^^ ^Q -> ^\ We specifically request more advice and information from the emacs community. END OF RATIONALE
ron@brl-sem.ARPA (Ron Natalie <ron>) (04/29/86)
I would recommend ^Q -> ^^ and ^S -> ^\ as this seems to be done for GNU, UNIPRESS, and Jove (which seem to be a large portion of the market). I would also like to endorse that ^@ (NULL) be grouped into the same category. There are terminals/systems that can not use NULL as an input character. We have gotten the JOVE and UNIPRESS people to add ESC-SPACE as an alternate binding to set mark. I would like to see this added to your proposal.
oz@yetti.UUCP (Ozan Yigit) (04/29/86)
Dick Grune & company posted a proposal for a "minimal" compatibility bindings, which, I think, is an excellent idea. (Oh brother.. now watch for the flooood in net.emacs !!! :-) Considering the number of functions Emacs & like editors have, plus all the exotic bindings that can be created, a minimal set of common bindings is the only way to ensure swithover from one brand of emacs to another without loss of productivity. I have, however, few reservations about how this set of bindings are determined: - various emacs & like editors considered, but it seems all had equal weight in the final set. If this is the case, very popular emacs bindings are effected by those that may not be as popular. My impression is that the bindings available in GNUemacs, GoslingEmacs & Jove should have more *bias* in the final set of common bindings than, say scame & microemacs. [Btw. I consider the Conroy version as the only true-microemacs.] - Other emacses *must* be considered. There are several other emacs-like editors in the market, along with others floating in the circuit: Elle and Montgomery's. Perhaps people on this newsgroup would provide the necessary binding information. I have not had the time to consider the proposed bindings, so nothing to say there. I hope a minimally complete set of bindings could be found. (Will it be called common-emacs ?? Perhaps anything that does not include the eventually-found-common-bindings should come with a warning from surgeon-general :-)) oZ -- The best way to have a Usenet: [decvax|ihnp4]!utzoo!yetti!oz good idea is to have a Bitnet: oz@[yusol|yuyetti].BITNET lot of ideas. Phonet: [416] 667-3976
leif@erisun.UUCP (Leif Samuelsson) (05/09/86)
In brief: Make a "Common Emacs Standard" independent of operating system. Ban C-S and C-Q. Keep one key for "Prefix Control". Keep M-[ (and maybe M-O) open for function keys. Let C-Z be a prefix. Being the author of one of the editors mentioned, I must say that I would like very much to see a common standard for key bindings. When I wrote Scame back in 1980, I had only RMS' original Twenex EMACS and the TOPS-10 look-alike called AMIS to compare with. These had the same key bindings, and since we used both these operating systems as well as Unix at the time, I chose to make as little deviations from this standard as possible. I think it is essential that the basic standard does not include anything that is operating system specific. There must be a common set of operators which work on any Emacs a user may choose to use on any operating system. Therefore, the use of C-S and C-Q must be banned from the basic set right from the beginning. I sort of like the idea of having C-_ be "I-Search Forward" and C-\ be "I-Search Backward" since C-_ is placed on "control-slash" on many keyboards and C-\ is placed on "control backslash" in many countries (not mine). I would also like a key reserved for "Prefix Control", since many keyboards can't send all control characters. As I am writing this article on my screwed-up Tele-Tec keyboard, I constantly use the sequence C-^ @ to set the mark. (Which works fine, but I can never get over the feeling of not having a C-D key). Scame does not have the function "Suspend" bound to any key as default. This is because Scame is restartable anyway, and many machines do not have job control. I personally use C-C for exit, C-X C-Z for suspend and M-+ for push to shell. In order to be able to map ANSI function keys, the standard should provide for the use of M-[ as a prefix. Since the popular VT100 terminal also uses M-O to start escape sequences, then maybe this should also be kept open. So what do we do with C-Z ? I say keep it as a prefix character. Allow the user to make up his own set of two-stroke commands from the vast set of unbound functions and macros available. ----- Leif Samuelsson enea!erix!erisun!leif (From July 1986: sun!lsamuelsson)
styborsk@hpspkla.UUCP (05/13/86)
If you are going to ban ^S/^Q due to their use in XON/XOFF flow control, you should also ban ^F due to its use in ENQ/ACK. ^F (ACK) gets eaten while ^E (ENQ) passes thru ok. This may be antiquated or of limited importance, but it exists. If you bind something to ^F, at least provide an alternative. Randy Styborski hpspkla!styborsk P.S.- How about banning character protocols in general? ;-)