> From: "Paul Guyot" <pguyot_at_kallisys.net>
> Subject: Re: [NTLK] Pauls future direction...
>
> The reason why SimpleMail and Newtscape are so slow is that they ask
> NIE to send them raw data as arrays of integers.
> So here we have the NewtonScript heap that suffers from a big array
> plus the fact that each byte (8 bits) is represented by a
> NewtonScript integer (32 bits in memory).
that's true for text -- there's some encoding mapping via arrays.
Newt's Cape xfers some files as binary (.pkg, .gif, .jpg, .pdb, etc.)
but that can be slow anyway, even when Content-Length is known and larger
buffers are used.
when there's no Content-Length provided by server, text resorts to
line-at-a-time transfer, and binary to byte-at-a-time(!) to avoid losing
data due to endpoint buffer bugs.
> Plus all the type check
> when you access it.
> It would be more convenient to have binaries there.
> Plus it would be better if the C++ code handled more of the transfer
> (like sending HTTP requests, checking the result, etc).
sounds good! (and maybe even time then to resurrect Waba;
or even better yet a 'real' Java... I've been spoiled by having nice
libraries for http:, https: connections).
-- Steve weyer_at_kagi.com Newton apps/tools: Newt's Cape, newtVNC, NewtDevEnv, Sloup, Crypto,... http://www.kagi.com/weyer/-- Read the List FAQ/Etiquette: http://www.newtontalk.net/faq.html Read the Newton FAQ: http://www.chuma.org/newton/faq/ This is the NewtonTalk mailing list - http://www.newtontalk.net/
This archive was generated by hypermail 2.1.2 : Thu Oct 31 2002 - 12:04:37 EST