• GophHub - Browse GitHub on Gopher

    From freet@freet@aussies.space (The Free Thinker) to tilde.gopher,tilde.services on Mon Jul 17 10:47:31 2023
    GitHub really annoyed me this month by starting to require
    Javascript in the web view of Git repos just to browse directories
    and source code.

    My answer is GophHub:
    gopher://tilde.club/1/~freet/gophhub/

    Still not completely finished, but the weekend's over so it is what
    it is. There are only two significant missing features left on
    my to-do list (plus line numbering in syntax-highlighted source
    files, which I've given up on (adding one excludes the other)).
    --

    - The Free Thinker | gopher://aussies.space/1/%7efreet/

    --- Synchronet 3.19a-Linux NewsLink 1.113
  • From xwindows@xwindows@tilde.club to tilde.gopher on Sat Jul 22 17:57:16 2023
    Glad you are being able to get back into Tildeverse Netnews now.
    Though I saw from the `Path:` of your post
    that you posted from Tilde.club;
    so I take that A. SSH forwarding on Aussies.space didn't work for this,
    and B. you finally solved the auth'signing algorithm issue
    in your normal client, or have worked around it.
    (How? If you don't mind telling)

    On Mon, 17 Jul 2023, The Free Thinker wrote:

    GitHub really annoyed me this month
    by starting to require Javascript
    in the web view of Git repos
    just to browse directories and source code.

    Yeah, noticed that for some time now.
    Interface-wise, it's now no better than the JavaS'creep-infected GitLab-- while GitLab is still a libre software.
    Now Gitea/Forgejo is way better in this regard,
    so as SourceHut, Gogs, Savannah, SourceForge/Apache Allura;
    or even barebone things like CGit, GitWeb, or Stagit
    in a specific browse-only department. [1]

    My answer is GophHub:
    gopher://tilde.club/1/~freet/gophhub/

    Thanks for writing this,
    it is indeed a needed workaround for plebs like us
    to deal with projects that somehow want to cling there for whatever
    (usually corporate, inertia, sunk cost) reason.

    Though it's still my stance that a proper long-term solution
    is advising people to get off that code-laundering (dis)service,
    refrain from participating (sending bug reports/patches) there [3],
    and/or avoid getting involved with projects which insist on such disservice
    to be a sole public participation channel in the first place.

    Regards,
    ~xwindows


    -----

    (WARNING:
    Rants ahead,
    and you probably already know most of this;
    so they are intended for as "to other people reading this reply")

    [1]
    To other people reading this reply: I will remind you
    that the original and proper way for an *external* developer
    to contribute a change to a public software project managed in Git
    is the way that is currently used in kernel Linux development [2];
    i.e. `git format-patch origin/master` and `git send-email`
    to the maintainer or project mailing list.

    (There is a very specific format of email message involved [4]
    which allows the maintainer to commit the change directly from MIME file. Easier alternative for less stringent projects
    [tilde-related projects counts in this category too]
    is one can simply write email to the maintainer
    to explaining what the fixes do,
    and attach the patch files produced from the first command)

    The "pull request" style of "contribution"
    (which needs infrastructure)
    is rather invented by GitHub to brainwash people
    to believe that such submission
    would inherently require on centralized service;
    in order to capitalize on network effect
    (i.e. you need account on single web service to submit changes).
    It is not even a natural way to maintain public participation
    in this SCM.

    [2]
    Kernel Linux is the original software project
    that Linus Torvalds wrote Git for,
    and the name "Git" was a reference to himself;
    supposedly for being stupid enough to fall
    for BitKeeper SCM's proprietary bait'n'switch ploy
    in spite of other people's warnings,
    and got burned by that in the end.

    [3]
    This do not imply that you cannot submit bug report or patches,
    if they publish contact email,
    mail those to them. [1]
    If they tried to redirect you to the disservice,
    you got declined,
    or treated with silence;
    simply share them as if you got responded WONTFIX on your report:
    post them the old way on your website,
    describe what they do,
    the copyright license you are publishing them under,
    as well as the exact HEAD/branch/repo
    that your patch series is intended to apply;
    optionally with your own reasons why they didn't get submitted
    on the "official channel".

    [4]
    When I submit change to a tilde project (like Tilde.club Wiki),
    I personally don't use `git send-email`,
    but rather craft my own MIME file manually
    from what `git format-patch` gave me,
    then `sendmail -t` the result (or import to Alpine and send);
    allowing me to have a discussion message in the front
    (usually me describing what the patch is and what it fixes/adds)
    and the patch content itself following right next to it in the message
    (rather than as an attachment) for a quick review,
    separated from each other by thin a scissor line "-- >8 --".

    Apart from maintaining one-mail-per-commit relation,
    this also allows the maintainer side to select Pipe command
    inside a TUI mail client to run `git am --scissors`
    and issue a commit right there if he approves.
    (I usually include an instruction too, whenever `--scissors` is needed)

    Different people might prefer different way to do this though:
    `git --format-patch --cover-letter`
    which produces an email thread is one of the other choices.
    --
    Contains ventilated prose,
    manually typed up in RFC 2646
    "text/plain; format=flowed" whitespace encoding;
    might read like poetry in some newsreaders.
    --- Synchronet 3.19a-Linux NewsLink 1.113
  • From freet@freet@aussies.space (The Free Thinker) to tilde.gopher on Sun Jul 23 23:48:58 2023
    xwindows <xwindows@tilde.club> wrote:
    Glad you are being able to get back into Tildeverse Netnews now.
    Though I saw from the `Path:` of your post
    that you posted from Tilde.club;
    so I take that A. SSH forwarding on Aussies.space didn't work for this,

    I didn't know that forwarding port 119 would allow posting. Hovever
    testing now in various environments, I get this error when I
    attempt to connect to the forwarded port from the Tin or Telnet and
    the connection is immediately closed:

    channel 2: open failed: administratively prohibited: open failed

    That error is from OpenSSH, PuTTY says nothing but has the same
    result (I wish I'd tried in OpenSSH first).

    and B. you finally solved the auth'signing algorithm issue
    in your normal client, or have worked around it.

    Actually for that I was connected via telnet to the SBC that I run encryption-insistent internet software on, and it was running
    OpenSSH 9.3. My current posting in Tin on the local spool via SSH
    to tilde.club is more of a one-off though - to laggy and awkward.
    I'll compile the latest Tin for the OS on that SBC one day and
    connect via NNTPS with that.

    Tonight I finally got a later version of PuTTY compiled for my old
    computers so I can now connect directly from them to tilde.club as
    well. It was a saga though, and failure to get a relatively recent
    CMake to compile means that I won't be able to upgrade PuTTY again,
    because later versions require CMake.

    As a result of that, I'm way too tired now and I'll respond to the
    rest of your post later (it's been a big day, I've been trying to
    heave half a ton of steel around, part of a very different weekend
    project to last week's).
    --

    - The Free Thinker | gopher://aussies.space/1/%7efreet/
    --- Synchronet 3.19a-Linux NewsLink 1.113
  • From freet@freet@aussies.space (The Free Thinker) to tilde.gopher on Tue Jul 25 17:48:50 2023
    xwindows <xwindows@tilde.club> wrote:
    B. you finally solved the auth'signing algorithm issue
    in your normal client, or have worked around it.
    (How? If you don't mind telling)

    For the next time SSH requirements get upgraded, I'm now
    considering attempting to cross-compile PuTTY by chrooting to a
    copy of the older distro's root file system on a PC running a
    current Linux distro, then using bind mounts and LD_LIBRARY_PATH to
    run cmake and make from the newer distro, thinking that they're
    running in the older distro.

    Right now I'm posting via SSH to tilde.club with PuTTY 0.76 (last
    PuTTY to build without CMake) running on a PC from the mid 90s and
    Linux from the mid 2000s.

    On Mon, 17 Jul 2023, The Free Thinker wrote:

    GitHub really annoyed me this month
    by starting to require Javascript
    in the web view of Git repos
    just to browse directories and source code.

    Yeah, noticed that for some time now.
    Interface-wise, it's now no better than the JavaS'creep-infected GitLab--

    Yes I was hoping to find a universal way to browse all Git repos
    without downloading their contents, so that it would work with
    GitLab as well, but Git just isn't designed for that unfortunately.

    while GitLab is still a libre software.
    Now Gitea/Forgejo is way better in this regard,
    so as SourceHut, Gogs, Savannah, SourceForge/Apache Allura;
    or even barebone things like CGit, GitWeb, or Stagit
    in a specific browse-only department. [1]

    Really just a directory tree of the Git repo indexed by the web
    server (with CGI processing disabled) would suit me. A separate
    tree with source code files syntax highlighed in HTML would be a
    bonus. Of course that's basically a static implementation of what
    GophHub offers via Gopher.

    My answer is GophHub:
    gopher://tilde.club/1/~freet/gophhub/

    Thanks for writing this,
    it is indeed a needed workaround for plebs like us
    to deal with projects that somehow want to cling there for whatever
    (usually corporate, inertia, sunk cost) reason.

    Though it's still my stance that a proper long-term solution
    is advising people to get off that code-laundering (dis)service,
    refrain from participating (sending bug reports/patches) there [3],
    and/or avoid getting involved with projects which insist on such disservice to be a sole public participation channel in the first place.

    Indeed, but if being bought by Microsoft wasn't enough to scare
    the open-source community off GitHub already, then I don't know
    what is.

    But as much as I dislike the centralisation around GitHub, when
    looking for obscure open-source projects or code examples by
    searching with Duck Duck Go (which admittedly is itself
    Microsoft-associated via Bing as a primary data source), adding "site:github.com" to the search string has become an extremely
    useful technique.
    --

    - The Free Thinker | gopher://aussies.space/1/%7efreet/

    --- Synchronet 3.19a-Linux NewsLink 1.113