The source is open. Please help.
https://github.com/isaacs/npm-www
Moving to google was only a "mistake" if you'd prefer a search that
literally does nothing whatsoever, except take the registry down every
few days.
The old
search.npmjs.org site used a technique that was not scalable
past a few thousand packages. At 12,000 packages, it finally was
unable to keep up with the task of building the search index, and
caused downtime and performance problems for the entire registry.
Keeping it as long as we did, that's the real mistake.
No one believes that the current search experience on the website is
ideal. But it is what we have for now. Please help with it. As much
as I'd love to spend every waking hour doing only this, I have many
other tasks to attend to.
In the meantime, the cli search is quite effective, and relatively
fast apart from the initial download of the index. But even that will
eventually fall over under the weight of Node's steadily rocket-like
increase in popularity. The only solution is to develop a good
server-side search solution that can be queried from the cli.
Microsoft has been kind enough to provide me with a free token for the
Bing API, so any solution that relies on that service would be pretty
awesome, and very easy to deploy and manage.
The source is open. Please help.