Re: [png-mng-misc] VOTE NO: URI keyword August 3, 2008

[Cosmin T. <co...@cs...>]

On Mon, 25 Aug 2008, Chris Lilley wrote:

> URI: hello world
>
> software is supposed to figure out that "hello" is a relative URI and "world" is a description?

Chris,

Neither the PNG spec nor any tEXt-supporting PNG encoder is able to 
impose any restriction on the text content input by the user. The user 
can type whatever he or she feels right.

Consider the following possibilities:

(1) "http://example.com"
(2) "http://example.com\n"
(3) "http://example.com/test1\nhttp://example.com/test2\n"
(4) "This is an example:\nhttp://example.com\n"
(5) "http://example.com/ My Very Important Bookmark"
(6) "hello"
(7) "hello world"

Out of these, (1-5) are likely to by typed by normal users, (1-6) are 
likely to be decoded by a decoder that follows the "URI" specification, 
and only (1) is likely to be decoded by the decoder that follows the 
"Source URI" specification.
That is why I argue that the "URI" definition is the useful one.

Besides, it's not much more complicated to make a _useful_ decoder that 
extracts the URI, but it is more complicated (at least in terms of UI) 
to impose a certain format that a _useful_ encoder will fill obliged to 
enforce, or at least warn, upon an incorrect entry from the unassuming 
user.

"Source URI" looks simpler, but it really isn't -- it just passes the 
buck on encoders, with an even heavier burden.

Best regards,
Cosmin