Re: [m17n-list] hi-itrans fix

[bd satish]

Hi,

I subscribed to the mailing list recently, so I haven't seen your
commits in December. On Linux Mint 14 (based off Ubuntu 12.10) neither
"ks" nor "ksh" are present. IMHO, whichever commit introduced those
mappings ought to be removed.

> I always prefer to type shankara to get शङ्कर rather than sha~Nkara.
And my preference is to type shankara to get शंकर rather than
shaMkara. See? It's not my preference or your preference that matters,
being as unambiguous as it can be, is all that matters.

> I have *never* needed to produce the character combinations क्श् ज्न  न्छ etc.

Just because *you* never needed doesn't mean others don't. Devanagari
is used for Marathi, Nepali, Magadhi, Maithili, Marwari, etc. Just
because क्श् doesn't occur in Hindi/Sanskrit doesn't imply other
languages don't have it. Think about the foreign phrases, say "Lux
soap" -> "लक्स् सोप्" . Such phrases may be rare, but they do occur.
Their frequency of occurence is immaterial for unambiguous mapping.

> I want a mapping which provides the most convenience to users, rather than 
> sticking to some dogma[s].

This is not my dogma, it's a mathematical property called prefix
coding [1]. No valid mapping in the system should be a prefix (start)
of any other valid mapping in the set. Sure, the aspirates (kh, ch,
gh, etc.) break this rule -- but this is the only case of ambiguity
and should not be a premise for introducing more ambiguities.

All of the four alternatives (a] to d])  involve more keystrokes than
needed for typing say न्ज्  which is currently simply " j + n ".

Having said the above, I'd probably wait for the other maintainers' or
users' opinion on the matter.

[1]: http://en.wikipedia.org/wiki/Prefix_code


--Satish


On 17 March 2013 22:04, विश्वासो वासुकेयः (Vishvas Vasuki)
<address@hidden> wrote:
> Hi Satish.
> On Sun, Mar 17, 2013 at 12:25 PM, bd satish <address@hidden> wrote:
>
>>
>> I'm able to type क्स्  just by typing "ks" even without your mapping.
>> The mapping ("ks" "क्स्") is simply redundant and not needed.
>
> Do you have the latest file? On my computer, due to the presence of ("ksh"
> "क्ष्"), ks does not map to anything. It has been that way since atleast
> last Dec, which is when I submitted many changes to these files.
>
>
>>
>> My guess is that you are trying to map Latin letters to Devanagari
>> glyphs, instead of mapping them to codepoints.
>
>
> True - I do plenty of typing in both kannada and devanAgarI. I want a
> mapping which provides the most convenience to users, rather than sticking
> to some dogma[s]. In other words, users should be able to type what they
> want using the most intuitive mapping, with the least number of key-presses.
> So, focus is strongly on users rather than dogmas.
>
> Please see below for examples.
>
>> In the same vein, all the below mappings are ambiguous:
>>   ("jn" "ज्ञ्")        --  then how would you type ज्न ?
>>   ("nch" "ञ्च्")    -- then how would you type  न्च ?
>>   ("nCh" "ञ्छ्")                            -- then how would you type
>> न्छ ?
>
> I always prefer to type shankara to get शङ्कर rather than sha~Nkara. I
> prefer to type sanchaara to get सञ्चार, rather than having to type
> sa~Nchaara.
>
>>
>> Next,  ("ksh" "क्ष्") is ambiguous -- how would you type क्श् then?
>
> Of course, I always prefer to type nakshatra for नक्षत्र and raakshasa for
> राक्षस. I would rather not have to press more keys than necessary, and
> 'shift' is one extra key press.
>
> I have *never* needed to produce the character combinations क्श् ज्न  न्छ
> etc.. As I said, I write a lot in multiple languages using both these
> scripts. In fact, I would request you to produce words which contain those
> characters from a dictionary of kannada, hindi or sanskrit. I doubt that you
> will find anything, and even if you do, I will be able to produce atleast 30
> times more words with the क्ष्, ज्ञ् ञ्च् glyphs.
>
> In the very same vein, for the earlier itrans mapping, one could have asked:
> "If ch is च्", how does one type "च्ह्"? Or ट्ह्? Or क्ह्?  So, it is not
> like these problems were newly introduced - it is inherent given the
> phonetic constraints of mapping latin alphabet to अक्षर-s.
>
> How did I type च्ह्? I have never needed to in all my writing, but if I need
> to here is what I do:
> a] c <space> <backspace> h
> b] c shift-space h <zero width non joiner>
> c] c <ctrl-shift-enter> h
> d] c <ctrl shift @ > h <zero width joiner>
>
> Everything except [a] only works with devanaagarii (hindi and samskrit is
> what i use). [a] works with kannada also.
>
> --
> --
> Vishvas /विश्वासः
>