On Tue, 4 Jun 2002, Richard Sitze wrote:

> Can someone give us a quick overview of when it is appropriate to HARDCODE
> CR/LF or LF in generated output, versus when we should use the system
> "line.separator" property?

The way HTTP is defined is 'on the wire'.

So you are not really sending out the text string (e.g. like a java
String) 'GET /something HTTP/1.0\r\n' - but rather sending out an octed
stream (i.e. a sequence of 8 byte entities in a specific left to right
order). Which just -happens- to look like human readable text when each of
those bytes are taken as indexes into the US-ASCII character table and
displayed in the order they where received from the wire.

Historically it was even stressed that the '/something' partial URI was
actually a direct capture of how you received the document in which it was
referenced in on the wire; i.e. regardless of actual encoding of the
document or the way the payload was transcoded or encoded within MIME.

This (unfortunately) was never fully understood or appreciated - giving
hours of fun now that we want to graft I18N onto a world where people take
a URI as a sequence of symbols rather than a sequence of bytes.

Dw

Reply via email to