[
Lists Home |
Date Index |
Thread Index
]
Claude L Bullard wrote:
> given your years of experience, why are multiple
> encodings ever necessary excepting the binary which
> I put in a class of its own?
Frankly, I think it is usually very hard to justify anything
more than one textual encoding and one binary encoding. Typically, the
textual encoding will be big, fat and full of compiler sugar that
makes it easy for humans to read. On the other hand, the binary
encoding should be as tight and fast to decode as possible for
whatever the underlying data type is. Other than that, the only reason
I can see for having more than two encodings would be to produced
encodings that are useful for signing. (i.e. like the CXER and DER
encodings). But, these "signable" encodings are usually best, and most
easily defined as profiles on more general encodings. You might also
have a need for variablity in your human readable encoding to handle
I18N (i.e. French folk don't like seeing thing like "street" or "zip
code" in an address record...) I18N also leads to all sorts of
"virtual" encodings in order to handle different character sets. (Like
the many versions of XML that differ only in character set.)
Typically, other than the two encodings (one for humans, one
for machines) discussed above, and ignoring the I18N-driven need for
variety in the textual encodings, the need for additional encodings is
usually just to respond to the damage done by idiots who have
inflicted upon the market some new encoding that contributes little
but still must be supported due to market acceptance. Sometimes they
do this because they are ignorant of what already exists. Sometimes it
is because they have a personal aesthetic preference for one style of
encoding over another, and sometimes it is because they intentionally
want to create a new encoding so that they can get some "proprietary
advantage" from using it. The reasons don't matter. Once a large
number of customers demand support for some stupid format-of-the-week,
you have little choice but to support it.
Whatever the case, I think you'll typically see much more
*need* for variety in textual encodings rather than in the binary
encodings. But, to answer your original question, I think that many of
the alternative encodings are simply unnecessary. But, it turns out
that the effort needed to support two encodings is pretty much the
same effort needed to support any number greater than one. Thus, it
makes sense to go the tiny extra step and provide support for any
number while doing one's best to ensure that there are really only
two....
bob wyman
|
