School of Biomedical Sciences, Medical School, University of Nottingham, Nottingham. NG7 2RD Sir, Lawrence and Giles [1] eloquently define the current problems with the World-Wide Web, but could "Nature" provide the solution ? The web has 800 million pages (and is still growing) but only 16% are indexed by the largest search engine. What is worse is that the newer search engines are biasing their index towards more 'popular' pages which may "delay or even prevent the widespread visibility of new high-quality information". The diminishing returns in indexing the whole web suggest that a large proportion of the diverse array of scientific information available on the web, from research group home pages to preprints and sequences, may never be discovered. Lawrence and Giles also point to a way forward: only 6% of web servers have scientific / educational content - a much more manageable amount to index. Metadata (information about information) is the key to better and more powerful searching, yet only 0.3% of servers use the Dublin Core [2] standard for metadata [1]. It seems clear that we need a science orientated search engine together with a scientific based set of metadata to help us trawl the oceans of information that could be available on our desktops. I think that a hybrid indexing scheme, halfway between the formidable task of a human generated internet catalogue like Yahoo, and the huge computer generated indexes like Alta Vista is the best way forward for a scientific index. The submission of an institutions home page would, after manual verification (thus excluding inappropriate sites), lead to the computer generated indexing of the rest of the site. The next step is a discussion by the scientific community to define suitable extensions to the Dublin Core metadata to describe the rich variety of scientific information available on the web. Nature's debates [3] would seem to be the ideal forum for such a discussion. Mike Gardner mike.gardner@nottingham.ac.uk References [1] Lawrence, S. and Giles, C.L. Nature 400, 107-109 (1999). [2] Dublin Core. The Dublin Core: A Simple Content Description Model for Electronic resources (1999); http://purl.oclc.org/dc/ [3] Nature. Nature Debates (1999); http://helix.nature.com/debates