## Motivation To get feedback, Comment System are essential to web pages. Most of modern web pages use 3rd-party services like Disqus. They are criticized by [ads] or trivial steps for sign-up/validation. For my Point of View, even using Github issue is not a good idea. I mean... You did make contents by yourself with markdown or some similar things, but why let others hold comments for you? [CGI] is another solution if you can access a machine with public IP. To show comments in your pages, you just need to transform HTTP Request into HTML. For example, [this][yasuoka-calendar] one-line comment button would be easily implemented! But consider spam or offensive content, a way to identify visitor still matters. ->Of course I am talking about Mail!<- For each comment, all you need to know are inside Mail header: sender title, mail address, datetime and SMTP server. By SPF/DKIM/DMARC, receiver can nearly 100% confirm Mail is from a specific sender. ## Mailing List Of course there is already a Comment System based on Mail: **[Mailing List]**. Actually I am not a mailing-list user (though I did subscribe some lists of [FOSS] projects). But thanks to public web page for archive[^1], I can realize how powerful it is. User just need to finish subscription, then everyone can start a new thread by mail, and get replied by another mail. But not everyone can build such this Comment System, you need to: 1. Be system admin to build related services 2. Tune HTTP web server to show archive 3. Afford traffic of SMTP for each new message in thread This might be a little overkill for basic Comment System on simple web pages. ## MDA and .forward If you can login as a normal user in Unix System (like [tilde.club]), `~/.forward` is what you need. From decades ago, Unix use program called [MDA] to deliver mail/message[^2] to local user (Most of time it is [Sendmail]). User can write their own `~/.forward` file to tell MDA where this mail should goes. For each entry in this file, it could be: 1. Another local user, for example: `alice`, `bob` 2. User in another host, for example: `alice@anotherhost.com` 3. Absolute pathname of a file, the mail would be appended to it. For example: `/home/alice/mails` Of course you can use `/dev/null` to drop any mail! 4. A command with leading char `|`, it would be executed with mail as STDIN. For example: `"|/home/alice/mail-handler.sh arg1 arg2"`[^3] With this approach, you can easily create a Common System based on mail: 1. Put a `mailto` link in your web page 2. Create a command to store incoming mail as comment in HTML format 3. Write an entry for this command in `.forward` 4. In your web page, display comments made by step.2 **Let's do it!** ## comment.sh Let's build a bash script called `comment.sh` for all of these: ### 1. Check mail is for comment I use pattern `Comment on page: ` in mail subject, to determine a mail should be stored as a comment. If it is not, script just return 0 safely: ```bash # RFC 2822 use Carriage Return, remove it for safety MAIL="$(tr -d '\r')" # join multi-line header field value into one line header="$(<<<"$MAIL" sed '/^$/ q; :a; N; s/\n\s\+//; ta')" # just mail body body="$(<<<"$MAIL" sed -n '/^$/,$ p' | sed '1d')" # determine mail is for comment by pattern pattern='^Subject: .*[cC]omment on page:? +(https?://[^/]+)?(/?[^ ]+).*$' <<<"$header" grep -E "$pattern" >/dev/null || exit 0 ``` ### 2. Determine output and text process tool For your HTTP routing, set `output_dir` to where HTML for comments locates. Also, I use `markdown_bin` to specify command to process markdown text in mail body ```bash output_dir=... markdown_bin=... ``` ### 3. Save each header field as variable - For easier usage, save each field key as upper-case, and replace `-` with `_` - Save datetime as rfc-3339 for more readable ```bash # enable execute last command in pipe under current shell shopt -s lastpipe; set +m; # save each field of header into variables echo "$header" | \ while read field value; do echo "$field" "$value" >>/tmp/header declare field=$(<<<$field tr [:lower:] [:upper:] | tr '-' '_' | tr -d ':') declare $field="${value}" done # formated datetime DATE=${DATE:+$(date --rfc-3339 seconds --date "$DATE")} ``` ### 4. Get path of output file Based on mail subject, determine path of output. For example: if mail subject contains `Comment on page: /posts/comment-by-mail.html`, then set output to `$output_dir/posts/comment-by-mail.comment.html` ```bash path=... output=$output_dir/${path}.comment.html umask 022; mkdir -p $(dirname $output) exec 1>$output ``` ### 5. Get comment from mail body By [RFC 1341], header with `Content-Type: multipart/*;` is separated by value of `boundary`. Some mail client automatically sends mail with both **text part** and **html part**. Here we extract text part if necessary: ```bash printTextPart() { boundary=$1 beginPat="\\|^--${boundary}\$|" endPat="\\|^--${boundary}(--)?\$|" sed -En "${beginPat},${endPat} { \@^Content-Type: text/plain@,$ {1,/^$/d; ${beginPat}d; p} }" } # check mail includes multiple part boundary="$(<<<"$CONTENT_TYPE" sed -En 's/^.*boundary="?([^"]+)"?.*$/\1/p')" if [ -n "${boundary}" ]; then # print mail part in MIME: text/plain body="$(<<<"$body" printTextPart ${boundary})" fi ``` ### 6. Write comment to output file Time to write mail body to output file. For Unix System like [tilde.club], set it readable for everyone: ```bash umask 133 ``` Then add basic ` LAYOUT fi ``` Finally, add a list item for comment: ```bash # Append STDIN after line number with variable, or after first