Currently, the spider could be making unnecessary requests, as it's only does conditional GET requests. While doing conditional GETs is a good thing, it should also avoid making requests by checking the Expires and Cache-Control headers to see if it should be relying on the cached version instead.

There ought to be a flag to disable this, if necessary.

Internally, it might be good to parse them into a single structure, converting max-age into a timestamp.

There's a chance that some feeds may not be sorted in the expected reverse chronological order. If a feed is refreshed, the order should be normalised.

One gap in the functionality of Planet Mercury over Planet Venus is the lack of a plugin/filter system.

I think this should be implemented in two ways:

1. A Go-based plugin system using the standard library's plugin package. This would be the most powerful method and would allow for the extension of the template system on top of giving ways to process entries. Having experimented with this, I'm not so sure it's a good idea. It requires the use of CGo. A better alternative would be to make use of net/rpc/jsonrpc and spin up small servers that can service requests to process entries. I'll need to work on what the interface will be, but initially, it'll probably be a single method called process that takes a serialised entry and returns a result along the same lines as the pipeline-based method.
2. An external method for processing entries. This would serialise entries as JSON to be read from stdin and read the result from stdout. The status could be used to determine whether the script was happy (0), wants the entry skipped (1), or is returning an error (2+, with the error serialised as JSON). I don't think it's a good idea to overload the semantics of the exit status like this now.

There's an opportunity for batching here, which could aid with filtering. I think the top level of the request and response should be an array of entries, even if the entries are just being passed through one-by-one. That way, if an entry should be skipped, omitting the entry will be enough.

Give a way to specify particular categories that should be included from the source feed. This would be useful for blogs that lack per-category feeds.

Themes should include a BOM indicating which files should be copied across to the output directory.

This is trivial to add. Mercury should publish an Atom feed.

See: https://github.com/marketplace/actions/goreleaser-action

When I create a release tag, I'd like to use goreleaser to generate the binaries and attach them to the release.

This was originally written in a bit of a rush, so I skimped on the test suite. I ought to fix that.

I believe this might be something to do with some particular RSS feed not including a published date or something:

panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x0 pc=0x3918bc]

goroutine 1 [running]:
main.feedQueue.Less(0x22bb200, 0x57, 0x80, 0x20f7200, 0x57, 0x80, 0x50, 0x4f, 0x64b01)
        mercury/feedqueue.go:35 +0x150
container/heap.down(0x4deed8, 0x209f6e0, 0x27, 0x57, 0x209f301)
        container/heap/heap.go:109 +0x104
container/heap.Init(0x4deed8, 0x209f6e0)
        container/heap/heap.go:46 +0x5c
main.main()
        mercury/main.go:95 +0x864

I'll investigate this, but I think the solution would be to filter out any bad feed items like this.