[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: encodings: do we need binary at all?

I don't think there's good cause to represent things in binary.

I think there's implementation evidence that complex encodings such as ber are
bad in that they are hard to implement and don't solve anything.

IP seems to have survived nicely by mandating a network order so that way
seems to work.

UDP datagrams aren't that small these days and in any case I think that kind
of point can be dealt with somehow.

keys and security and the like is hard enough to deal with that making it
more complex by adding asn.1 and ber or der is not worth the extra pain, in
my opinion.

and anyone who wants to bark at me about how silly it is to question ASN.1
needs to send me the address of their anonymous FTAM site if they want me to
believe that an ISO spec from ten years ago is somehow intrinsically a good
thing ;-)

I'm delighted to refer to the LEGACY ISO-based stacks I have to deal with
when I claim I've a bit of experience in this...


At 03:28 PM 2/24/96 -0500, you wrote:
>
>There is an interesting argument being brought up here -- are binary
>encodings of any sort strictly needed?
>
>Recently, I was playing with a colleague with a strawman certificate
>format for which we developed a very trivial ad hoc binary encoding,
>which was not too onerous to use.
>
>On the other hand, it has occured to me that given the general
>overhead of public key systems, binary formats might not be needed at
>all. They probably don't provide a performance advantage. The only
>advantages they provide are some increase in compactness and some ease
>of producing a cannonical encoding for signature purposes. Of course,
>one can think of ways of achieving the latter without binary
>encodings, and the former isn't necessarily such an important feature.
>
>To give a really extreme example, this is an ASCII representation of a
>public key from the SSH system. It was originally just a single line,
>but I've broken it up with backslash/newlines to make it more readable:
>
>1024 37 85799595668887144041234497858720965191494394274243010276995452\
>7289397549856435807651462408205171548319480002414448720219295632456503\
>7288575015538644624806921929027414093682073278113060408377555754422808\
>5989367574906462469630862751348976715281503899082491996857682292734325\
>009646270999199065958155229155773963 perry@jekyll.piermont.com
>
>Now, this isn't a certificate -- its just a key (the last field
>containing the name@host is just a comment) -- but one wonders a bit
>about why representing these things in ASCII is such a bad thing.
>
>The biggest problem in my mind is the fact that ASCII makes it harder
>to send around certificates in small spaces like UDP datagrams, but
>again, using bases other than 10 for the numbers alleviates that
>somewhat and it is also not clear how important this compactness is in
>the first place.
>
>Comments?
>
>Perry
>
>

                  Rodney Thayer           ::         rodney@sabletech.com
                  Sable Technology Corp   ::              +1 617 332 7292
                  246 Walnut St           ::         Fax: +1 617 332 7970     
                  Newton MA 02160 USA     ::  http://www.shore.net/~sable
                           "Developers of communications software"