Hello,



I got questions, thus I try to explain it a little more verbose:



On Tue, May 30, 2006 at 05:56:49PM +0200, Heiko Oberdiek wrote:



> On Thu, Apr 27, 2006 at 08:22:51PM +0200, Heiko Oberdiek wrote:

>

> * I have written a wrapper file that merges all the ltnews??.tex

>   documents: ltnews.pdf contain all LaTeX News issues 1--17.



Now base-tds.zip contains

    doc/latex/base/ltnews.pdf

instead of

    doc/latex/base/ltnews01.pdf

    doc/latex/base/ltnews02.pdf

    ...

    doc/latex/base/ltnews17.pdf



(The source/latex/base directory is not changed,

the files ltnews01.tex, ltnews02.tex, ... remain there.)



> * To get smaller pdf file sizes I have experimented with

>   destinations:

>   * Unused destinatins are removed.

>   * The destination names are renamed to shorter names.

>     I am using all bytes (0-255) except for

>     - 0: dangerous for PDF viewers that work with C strings (xpdf?)

>     - 13: carriage return needs quoting, otherwise it would be

>           normalized to 10 (line feed).

>     - 40, 41, 92 (parentheses, backslash): needs quoting

>     - 255: avoids unicode marker at string start

>     I would be interested, if there are problems with the links/outlines.

>     (Apart from links that unhappily point to the baseline.)



Consider an HTML page foo.html with a lot of internal links:

    <a name="section.1">...</a> or <a ref="#codline.127">...</a>

In a similar way destinations are used in the PDF file.

Additionally the names are also stored in a data structure with the

function of a search tree.



The correct spelling of the destination name does matter, if someone

wants use the destination externally: http://somewhere/foo.html#section.1

But in case of the documents that I have generated for latex-tds, these  

destination names are (randomly) choosen by hyperref, not a property     

of the document. You cannot rely that the destination name for the       

implementation section is "section.4", it can be "section*.27" or        

"chapter.3" or ...



Thus I made the assumption the spelling of the destinations are not

intended for external referencing. This allows two optimizations:  

* Unused destinations are removed.

* The destination names must still unique, but I can choose shorter

  names, e.g. instead of

      section.1, section.23, codline.127   (30 bytes)

  the shortest possible, but unique names are used:  

      A, B, C                                (3 bytes)

  The HTML example from above would be transformed to:

      <a name="A">...</a> or <a ref="#C">...</a>



Example: file size decrease of source2e.pdf: around 4.5 %.



Yours sincerely

  Heiko <[log in to unmask]>