Who need a CMS, anyway?

E-mail to the rescue! I believe others already did something similar.

The gemini protocol is fine and dandy. However, it doesn't, directly, support the upload of resources, making it non-trivial to create new content in an easy and portable fashion.

I was thinking for a while how to solve this problem for myself, given that, until today, I used to just store my gemini articles in a fossil repository, with a touch of "make" to build the capsule and it's web counterpart. This approach has some disadvantages:

• It depends on a bunch of installed tools for developers (fossil, make, etc.).
• It requires you to, for sure, be in front of a computer to publish an article, or to make any change (no portability).

Looking at the gemini landscape in particular, the only solution seems to be to implement the titan protocol.

=> The titan protocol.

However, I really, really did not want to write yet-another-daemon, or change gmnxd to handle this specific case. Not convinced, I started from the base, writing down the following requirements:

• The solution must be device agnostic (being able to write from the phone, my pc, a tablet or even a guest computer).
• At a minimun, it must be able to handle the publishing of an article and the upload of related resources (files, images, etc.).
• Publishing a gemlog, should automatically mean that the web counterpart is generated.
• In general, I must NOT need to directly login into a machine for a shell or whatever.
• Write the less amount of code possible and leverage the existing infrastructure.
• If possible, do not run another service!
• No need to fiddle with paths directly. Write, push a button and see the magic happen.

After some heavy thinking, I understood that, what I needed, was something similar to a mailing list manager, but for my gemini capsule. Leveraging SMTP would let me to use any device where I have an e-mail account for my domain setted-up. In the end, e-mail clients are everywhere. Mapping my desiderata to an e-mail was quite easy then:

• The user part of the destination e-mail addresses, identifies one of my capsules.
• Commands: they are defined in the subject, in a key:value format (i.e. new:My new article creates a new gemlog with that title).
• In the e-mail body, the text/plain part represents the gemtext content, text/html is discarded.
• E-mail attachments, generally of type application/octet-stream, are additional resources that must be uploaded with an article (images, files, and whatnot).

My idea was to leverage, potentially, whatever e-mail client du-jour to publish an article whenever I wanted. In essence, whatever tool I would have written, needed to be able to take an e-mail, extract it's subject and destination and parse the MIME part in it's body. And that's exactly what I did, by writing a small custom MDA (Mail Delivery Agent).

An MDA is nothing more than a program that receives a full e-mail message in it's standard input. The MDA I wrote is composed by two parts, a shell script driving the majority of the automation and a TCL script leveraging the mime package to extract all the mime parts from the e-mail body.

=> The mime package

The steps are:

• E-mail is received and stored by the MDA in a temporary file.
• The command is extracted from the "Subject" header.
• If the command is to create a new article, a directory, if non-existent, is created for the current day (path/YYYY/MM/DD format).
• The title is extracted from the subject also.
• The e-mail body is parsed and all the mime parts are saved in related files.
• The gemtext is put in the right directory and the file properly named, along the other attachments that were present in the e-mail.
• Gemtext converted to HTML through gmi2html and the same process is done the the WWW site.
• Indexes for the blogs of both the versions (Gemini and WWW) are updated.

In the end, I implemented the following commands so far:

• New article publishing.
• Updating of an existing article.
• Uploading of a generic file.
• Deleting of a resource.

And everything is done for both the gemini and WWW versions of the site.

Security

To avoid potential issue with site hijackers, two things were modified in the opensmtpd configuration: a specific rule for the new MDA was put into place permitting only to a specific e-mail address to write to the gemini addresses. And that same rule action adds the authentication informations to the "Received" headers. This way I can, in the MDA, check if the e-mail was sent from a specific, authenticated account, discarding anything non-authenticated.

How does it works, in practice?

Well, better than expected. Leveraging the whole E-mail ecosystem (from the clients, to the SMTP protocol) on top of my automation, makes everything smooth. Combined with the fact that I have an IMAP server, it means that I can have my articles saved as "drafts", or, by putting myself in CC, manage the "revisions" by leveraging the threading capabilities of any e-mail client. I'm still exploring how to leverage this whole e-mail thing more, but it's already good enough to have a simple client to write, attach pictures and what not and with a push of a button, have your article published.

P.S.: No, I will not publish the code. It's tailored to my needs and hacked together. I'm not proud of it's quality.

In the end, Bananas!

=> bananas.jpg

If you want to know more, feel free to contact me in private, or on IRC, or the fediverse.

Enjoy!