This has been achieved by spawning off separate Unix processes to deal with each of the input'ed or include'd files. As each process terminates all the space that it used is reclaimed. Asynchronous communication between processes takes place using the Unix DataBase Management system (DBM or NDBM) which should be present on your system.
To take advantage of these changes, it is necessary to split the source text into more than one files which can be assembled using the LaTeX input or include commands.
Two new options -no_images and -images_only allow ``off-line'' image conversion. The advantage of using these options is that the translation can be allowed to finish even when there are problems with image conversion. In addition it may be possible to fix manually any image conversion problems and then run LaTeX2HTML again just to integrate the new images without having to translate the rest of the text. More instructions on how to do this are included in the ``Troubleshooting'' section of the LaTeX2HTML manual.
Can now use either the pbmplus or the netpbm libary. If netpbm is used then it is no longer necessary to get and install giftrans in order to generate transparent inlined images.
Also, a new option map=<image map URL> in the command htmlimage can turn an included postscript image into an active image map.