1Using browser and the Internet/the Web

1.1Removing nag screens

1.1.1How to use this

Copy the respective fragments to your browser's JavaScript Console (Ctrl+Shift+J on Chromium).

Don't run codes you don't trust.

1.1.2Make www.webtoon.com fast

That website has an unacceptably slow scrolling. This script makes it fast.

  1. Usage

    Open the comic episode you want to read.

    Paste this fragment into your browser's JavaScript console.

    // Retain big images. Discard everything else.
var images = [];
document.querySelectorAll("img").forEach(function (element) {
    images.push(element);
});
document.head.innerHTML = "";
document.body.innerHTML = "";
images.forEach(function (element) {
    const big = 256;
    const url = element.dataset.url;
    if (url && element.width >= big) {
        element.src = url;
        element.style.display = "block";
        document.body.appendChild(element);
    }
});

    To go the the next episode, increment the episode_no parameter in the address bar.

  2. Notes

    I tried getEventListeners and removeEventListener but they don't work.

1.1.3Remove Quora nag screen

Tested on 2017-07-08.

Not only does Quora put up a nag screen, it also disables scrolling.

document.querySelectorAll("div[id]").forEach(function (x) {
    if (x.id.indexOf("signup_wall_wrapper") >= 0) { x.remove(); }
});
document.body.classList.remove("signup_wall_prevent_scroll");

1.1.4Remove Pinterest nag screen

Tested on 2018-05-13.

Type 1.

document.querySelector("[data-test-giftwrap]").remove();

Type 2.

document.querySelector("#desktopWrapper").style.position = "static";
document.querySelector("body style").remove();
document.querySelector(".FullPageModal__scroller").parentNode.remove();

1.1.5Group Gmail mails by sender email

(function () {
    var group = {};
    document.querySelectorAll("[email]").forEach(function (elem) {
        var sender = elem.getAttribute("email");
        group[sender] = 1 + (group[sender] || 0);
    });
    var list = [];
    var sender;
    var count;
    for (sender in group) {
        count = group[sender];
        list.push([count, sender]);
    }
    list.sort(function (a, b) {
        return -Math.sign(a[0] - b[0]);
    })
    return list;
})();

1.1.6Chrome bookmarklet: Make Markdown link for page

This may produce invalid Markdown. Check before you copy.

javascript:window.prompt("Copy to clipboard: Ctrl+C, Enter", "- [" + document.title + "](" + document.URL + ")");

1.1.7tvtropes.org: show all spoilers

1.2Searching the Internet

<2019-02-04> For reverse image search, I find Yandex better than Bing, and Bing better than Google.

  • Example search: given a text description, find related images
  • Example reverse search: given an image, describe it, or find its copies on the Internet

  • Internet search tools in 2018

    • duckduckgo.com: privacy-focused search engine

      • less censored than Google
      • no reverse search (yet?)

    • google.com: biggest Internet search engine in 2018

      • censored
      • text search
      • image search
      • reverse image search

      • Problems

        • Google should rank down Pinterest.

          • In my experience, Pinterest is never an authoritative nor original source.
          • Google has ranked down Tumblr and Wikipedia.

        • Example of Google censorship (compare with DuckDuckGo):

          • search terms related to pornography
          • search terms related to the darknet (the hidden wiki, Tor websites)
          • rhino poaching (Google favors contents against rhino poaching)

        • Richard Stallman's reasons not to use Google

    • bing.com: Microsoft's answer to Google

    • searx?
    • startpage?

    • reverse search tools

      • Google and Bing have reverse image search
      • tineye.com: reverse image search

      • moomash (was "audentifi"), YouTube reverse audio search

        • 2018: only works with some video formats; doesn't work for old videos
        • 2018-09-15: It have gone out of business, it seems.

      • Shazam

        • Is there anything like Shazam that takes a YouTube URL, and doesn't require me to install anything?
      • http://whatpixel.com/original-source-image-search-tools/

      • saucenao.com: reverse image search; I use it to find images stolen from pixiv

  • The Internet is full of crap.

    • So are libraries.
    • So is the world.
    • So is this website.
    • Thus, you must think for yourself.

1.3Contributing to Wikipedia

1.3.1Beginning contributing to Wikipedia

Don't publish anything you may regret later. Once it's on the Internet, it's forever. The Internet doesn't forget. Wikipedia is a highly crawled and archived site.

1.3.2What should be in user page?

Templates: User page, User in $country, Babel

See WP:User pages.

WP:Userboxes

1.4Reading Reddit in Indonesia

  • Read Reddit in Indonesia with Tor Browser.
    • Pro:
      • Very easy to set up, very low friction.
    • Con:
      • Some speed penalty.
      • Should not log in.

2Making a personal wiki

3Ansible

An inventory is a map from host alias to host address. We use those aliases to select the machines to mutate.

Ansible has two executables: ansible and ansible-playbook.

The ansible-playbook takes a YAML configuration.

The ansible executable executes one command.

The ansible command is like a "single-task playbook".

See also man ansible ("run a task on target hosts") and man ansible-playbook.

A machine can have many roles.

An Ansible role should be a noun phrase (web-server), not a verb phrase (install-web-server).

4Amazon Web Services

5Moving to clouds, for old-school sysadmins

The most important pages on the AWS website are the pricing pages and the technical documentation. That website has much content, but not much information, perhaps because they are not selling to sysadmins.

Amazonese Old-school
Route 53 managed DNS server
VPC (virtual private cluster) managed LAN (local area network)
EC2 (elastic compute cloud) instance managed virtual machine
security group managed iptables/firewall
RDS (relational database service) managed SQL server
EBS (elastic block store) managed NAS (network-attached storage)
ELB (elastic load balancer) managed HAProxy
ElastiCache managed Memcached/Redis
Lambda automatically turn on machines to run a piece of code, and turn off idle machines

AWS, GCE, and Azure do the same thing you used to do. The difference is they do it on a much larger scale, and they make an API on top of it, so you can automate it, but this also mean that you can be automated away, so beware!

With this cloud stuff, you can't buy a machine and bring it to the data center. You start a machine from your computer. The machine is now virtual; it doesn't correspond to a motherboard anymore. Procuring a machine is just a few clicks on the website, or a few keystrokes on the terminal, and your machine will be running in a few minutes. What you used to do in days, now you can do in minutes.

With this cloud stuff, you can't visit the data center to restart a stuck machine. You restart it from your computer.

You're billed per hour. What was infrastructure (like roads) is now utility (like electricity).

The cloud is cheaper for bursty load with low average load. If your average load is high, old-school is cheaper.

One thing doesn't change: you still need to back up data to a safe place outside the cloud. (I'm a hypocrite; I say that but I don't do that.)

6Deploying web applications

6.1Formalizing deployment requirements

What is the way to deploy web applications?

  • General information

    • I have a Java web application.
    • It compiles by mvn package.
    • Its main class is blah.

  • Network

    • It listens on port 1234.
    • Its URL should be https://blah/.
    • It is HTTPS only. HTTP port shouldn't be open at all.

  • Resource requirements and burst characteristics

    • It needs 4 GB of RAM for acceptable garbage collection overhead.
    • It is mostly idle, but when it bursts, it requires 4 cores.
    • Ops is free to horizontally scale the stateless application server.

Assuming that I'm on either Amazon Web Services or Google Cloud Platform, how do I formalize my ops requirements in a cloud-agnostic way? The 2016 article "On Formalizing and Identifying Patterns in Cloud Workload Specifications" paywall suggests an answer:

  • "Approaches include orchestration specifications CAMP [1], [2], Open-CSA, SOA-ML and USDL, and on the industrial side solutions such as Amazon CloudFormation, OpenStack Heat, Cloudify and Alien4Cloud. To consolidate and enable interoperability within this variety of approaches, a technical committee by OASIS [3] defined a standard for the Topology and Orchestration Specification of Cloud Applications (TOSCA) [4], [5], which defines guidelines and facilities for the complete specification, orchestration and configuration of complex workloads, addressing portability in heterogeneous clouds."

I assume that it suggests TOSCA.

Problem: The average person won't read a specification.

Who uses TOSCA? Who implements that? Why do I never see it on AWS or GCP? Why would they follow your standard if they are the de facto standard?

Do we need more than one cloud providers?

Tools?

Why I don't use NixOS:

  • NixOS is insane. Patching every software on Earth is not sustainable. NixOS is only sustainable if the upstream developers use NixOS.
  • 2018-08-31: Ubuntu has old-releases.ubuntu.com. It archives things back to 2006. Ubuntu has money to host 12 years of archive. NixOS doesn't have that much money. NixOS can only afford to host 1-2 years of archive.
  • NixOS (or anything else indeed) would be heaven if library writers valued backward compatibility. I want my library writers to worship backward compatibility like sysadmins worship uptime. I want them to never break things that depend on them. But my experience invalidates this hope. I've seen too many breakages.

Who uses this?

  • 2013 "Towards a Formal Model for Cloud Computing" paywall

6.2Design

I agree that 2018 devops ontologies suck, but I think we shouldn't avoid ontology-based systems. The solution is not to avoid ontologies. The solution is to craft a proper ontology that is timeless and essential. This is a hard philosophical problem.

For example, the relationship between "application" and "entry point" is timeless. "Entry point" is an essential property of "application". By definition, every application has an entry point.

Every software has an implicit ontology, like it or not.

Every ontology systems that captures accident instead of essence is bound to fail. Every computer ontology system that avoids philosophical ontology (What is X? What is the timeless essence of X?) is bound to fail.

Are these related?

How do we answer "What is X?"?

There is an easy answer for mathematics. Mathematics is unique in that its ontology is mostly a priori / by fiat: we say it exists; therefore it exists. However, would it still be the case if we didn't have languages to express it?

For the real world it's hard.

Sometimes when we ask "What is X?", we are really asking "What is X for?" instead.

How do answer "What is X?" such that the correctness/truth/relevance of the answer does not depend on time/circumstances? We don't know how to predict the far future.

6.3Continuous something

  • Continuous integration

  • Continuous delivery

  • The ideal workflow: Git push triggers deployment?

These pages may be outdated:

6.4Fabric vs Ansible

From the user's point of view, Fabric is a python library, whereas Ansible is YAML-driven framework.

6.5What is DevOps?

XML is not suitable for declarative DevOps. See comments in devops/example.xml. We want the declaration site of some bindings to be as close as possible to their use sites.

I'm thinking about using Dhall.

Separating Dev and Ops doesn't make sense.

Ops can't fix shitty Dev. No amount of Ops will fix stupid programming. Ops is impossible without decent Dev.

What is Google search result for "devops tools"?

6.6Haskell for devops?

6.7Migrating running processes

6.8Troubleshooting Dashboard: What metrics you should monitor and why?

  • We want to minimize what we need to see. we want the metric that predicts the most problems.

    • Rising maximum latency is a sign that something is overloaded.
    • Rising resource usage (CPU, memory, disk, disk queue depth) predicts rising latency.

  • We want the metric to help us locate problems.

When such metric deviates from baseline, we know there is problem, but where?

What other metrics should we monitor?

  • HTTP 4xx and 5xx status codes and connection failures?

7Using Java

<2019-01-30> Here are some Java stuff. I used Java back in the days when I lacked self-respect. I recommend Prolog instead of Java.

7.1Profiling?

7.2Using Maven

The strength of Maven is its ecosystem. Network effect.

Maven is rigid. All projects build in the same way. If you want Maven to do something, but you can't find an example on the Internet, you should assume that it can't be done.

To compile your project, run mvn compile.

To package your project, run mvn package.

Maven output directory is target.

7.3Using Gradle

  • Conclusion:

    • I still haven't found any reason to switch from Maven to Gradle (other than "because this project is already using it").

  • Which version of Gradle are we talking about?

    • Gradle 2.9.

  • What problems does Gradle solve?

    • Dependency management (picking the libraries' versions and downloading the corresponding JAR files) for Java.

  • Do we have those problems?

    • Yes. Software has external dependencies.

  • Does Gradle 2.9 solve those problems well?

    • No. Gradle 2.9's dependency resolution algorithm doesn't compute the intersection of version ranges. (Maybe now it does.)

    • However, there are times we want exact versions instead of version ranges. You want deterministic builds. In this case, there's no need to compute intersections.

    • On the other hand, we want to benefit from library updates. Maybe there are security fixes. So we want version ranges?

      • But this assumes that the library maintainer obeys Semantic Versioning.

  • Why are we using Gradle instead of Maven? What Maven annoyances does Gradle hide from us?

    • Gradle build scripts are shorter (but IDEA autocompletes Maven pom.xml).

  • Why not Gradle?

    • Why Maven instead of Gradle?

      • IDEA integrates better with Maven because pom.xml is configuration, not program. (But IDEA can also open build.gradle?)

        • Opening a pom.xml just works in IDEA.

  • How does Gradle 2.9 annoy us?

    • We often have to explicitly tell Gradle 2.9 the exact versions of the libraries we want because it doesn't compute intersections properly. Maven computes intersections.
    • Gradle doesn't download dependencies in parallel. (But neither does Maven.)
    • Gradle 2.9 doesn't generate a useful Maven POM, only a minimally valid POM.

  • When should I split a Gradle subproject or a Maven module?

    • When we need to reuse one subproject without the others.
    • If they don't make sense separately, don't split them; it'll just slow down the build for nothing.
    • The same goes for Maven modules.

  • Woes

    • ShadowJar doesn't work with Gradle 2.13.

7.4<2017-02-10> Make Open Travel Alliance XML schema work with JAXB

Save this to a file named like BINDFILE:

<bindings xmlns="http://java.sun.com/xml/ns/jaxb"
          xmlns:xsi="http://www.w3.org/2000/10/XMLSchema-instance"
          xmlns:xs="http://www.w3.org/2001/XMLSchema"
          version="2.1">
    <bindings schemaLocation="xsd/OTA_CommonTypes.xsd">
        <bindings node="//xs:attribute[@name='Value']" multiple="true">
            <property name="ValueAttribute"/>
        </bindings>
    </bindings>
</bindings>

Then run:

xjc -b BINDFILE -d outputDir xsd/XSDFILE

7.5Production command-line

<2018-08-29> For Java enterprise/server programming, I recommend Maven instead of Gradle.

This is a command line for production. Replace variables accordingly.

The GC log is important. When your application is unexplainably slow, first look at the GC log: does GC take too much time? Look at the "real" time.

8Using Python

Python virtualenv is relatively forward-compatible. Don't waste time installing Python from source. Use the Python packaged with your distro, and use virtualenv. The Pip that comes with Python 3.7.0 fails because Ubuntu 14.04 OpenSSL is too old (or Python doesn't bother to maintain backward compatibility).

sudo apt-get install python-virtualenv

Create a virtualenv directory using virtualenv PATH Note: After the directory is created, it can't be renamed.

How does python find packages?

9Software

10Using Haskell

10.1Setting up development environment

10.1.1My incoherent rambling

My old way: stack. My new way: cabal new-style. It may change again.

  • Install the Haskell stack tool.

    • I use the manual download.

      • If you want to make sure that the download isn't corrupted, check the corresponding sha256sum from GitHub releases page.
      • If you don't mind sudoing, use the installer script in the documentation.
    • Then I check the archive contents using tar tzf.
    • If there's no weird paths, I extract the archive with tar xzf.
    • Then I make the symbolic link ~/.local/bin/stack.

    • If you use the manual download, you may have to install some operating system packages.

      • The list is on the install script. Search for install_dependencies for your distro.

  • Choose a Stack solver.

    • Forget it. Just install GHC to home.

      • ./configure --prefix ~/.local
      • make -j4 install

  • Which version of GHC should I use? - The one that is supported by HaRe (Haskell refactoring tool) and Leksah. - On 2018-08-20, this is 8.0.2. - Leksah requires ghc >= 8.0.2. - HaRe supports ghc <= 8.0.2. - GHC 8.0 is unacceptably slow. - Forget HaRe. We'll go with Leksah. Use GHC 8.4.3.

    • The widely supported GHC version lags very much behind the latest stable GHC version. I think this may be because the GHC team is rolling out lots of breaking changes in the parser because they are working on the "Trees that grow" proposal.

    • Which Stackage LTS version should I use?

      • LTS 6.35 if GHC 7.10.3?

        • It also hosts a hoogle search for searching Haskell program elements.

  • How to get started?

    • Too many choices

      • install using the package manager that comes with your system

        • pros: least hassle
        • cons: outdated software
      • Stack
      • Cabal
      • Nix

      • Haskell Platform

  • What is your preferred way of installing Haskell?

    • Install cabal-install

    • Install current stable release of GHC

    • Modify PATH in ~/.basrhc:

      • Ensure that the line export PATH"$PATH:$HOME/.local/bin"= is in ~/.bashrc.

    • For what is the hassle?

      • So that, if anything goes wrong, I can nuke it without nuking my whole operating system.

10.2Hackage outages

Hackage is Haskell package repository. Sometimes it goes down.

10.3Haskell in 2018

10.4Using GHC

10.5what

10.6Finding a Haskell IDE

I haven't found a convincing IDE for Haskell.

10.7what

10.8what

10.9Curating libraries

10.10what

10.11Haskell woes

10.12GHC woes

  • Profiling requires recompiling all transitive dependencies if they happen to be compiled without profiling.

10.12.1Working on GHC

10.13People who have too much time

10.14What's hampering Haskell adoption?

  • GHC's aggressive intermodule optimization precludes prebuilt binaries.

10.15Using Cabal

  • Every package used by Setup.hs must have a vanilla version.

    • Why I encountered this error:

    • How I encountered this error:

      • HDBC-postgresql-2.3.2.5 Setup.hs build fails due to linking error with shared: True.

        • The offending packages are old-time and old-locale.

          • It's OK if we build them as shared library, but we must also build their vanilla version.
        • It's not the library content that fails to build. It's the Cabal Setup of the library that fails to link.
        • GHC expects that old-time and old-locale are system libraries?
        • GHC passes '-lHSold-time-VERSION-HASH.so'. It passes that in the -l switch to GCC.

        • Cabal writes 'libHSold-time-VERSION-HASH-ghc-8.2.2.so'.

          • Note the ghc-8.2.2 part isn't in the string passed by GHC to GCC.
        • Can we solve this by cabal install old-time old-locale?

        • https://github.com/haskell/cabal/issues/1720

          • Workaround: Add --ghc-options-dynamic= to cabal new-install
        • How do we tell Cabal to use the version we installed with new-install?
        • Where should we fix this? Cabal? GHC? HDBC-postgresql?
        • Should we find another library? Hackage has a low-level libpgsql wrapper.

        • Should we just disable HDBC-postgresql on meta?

    • How I diagnosed it:

      • Pass -v to GHC (create a bash script named ghc that calls SOMEWHERE/ghc -v "$@", and put its directory in front of PATH).
      • Add -v to cabal. Then look at this fragment. It's suspicious that old-time is the only package with a hash.
      package flags [-package-id Cabal-2.0.1.0{unit Cabal-2.0.1.0 True ([])},
     -package-id array-0.5.2.0{unit array-0.5.2.0 True ([])},
     -package-id base-4.10.1.0{unit base-4.10.1.0 True ([])},
     -package-id binary-0.8.5.1{unit binary-0.8.5.1 True ([])},
     -package-id bytestring-0.10.8.2{unit bytestring-0.10.8.2 True ([])},
     -package-id containers-0.5.10.2{unit containers-0.5.10.2 True ([])},
     -package-id deepseq-1.4.3.0{unit deepseq-1.4.3.0 True ([])},
     -package-id directory-1.3.0.2{unit directory-1.3.0.2 True ([])},
     -package-id filepath-1.4.1.2{unit filepath-1.4.1.2 True ([])},
     -package-id ghc-prim-0.5.1.1{unit ghc-prim-0.5.1.1 True ([])},
     -package-id old-time-1.1.0.3-8c2cc8e5fb3b424e71501141225064c5d9ee4eeba7f40b702227ad1c3ea2c5b7{unit old-time-1.1.0.3-8c2cc8e5fb3b424e71501141225064c5d9ee4eeba7f40b702227ad1c3ea2c5b7 True ([])},
     -package-id pretty-1.1.3.3{unit pretty-1.1.3.3 True ([])},
     -package-id process-1.6.1.0{unit process-1.6.1.0 True ([])},
     -package-id template-haskell-2.12.0.0{unit template-haskell-2.12.0.0 True ([])},
     -package-id time-1.8.0.2{unit time-1.8.0.2 True ([])},
     -package-id transformers-0.5.2.0{unit transformers-0.5.2.0 True ([])},
     -package-id unix-2.7.2.2{unit unix-2.7.2.2 True ([])}]

    • How I solved it:

  • #4506 new-haddock's file monitoring broken

    • new-haddock doesn't work after new-build.

  • #5290 new-build runs into internal error after deleting from store

    • Solution: nuke the entire store: rm -r ~/.cabal/store.

  • Non-problems

    • cabal new-build --disable-optimization doesn't disable optimization of transitive dependencies.

      • Cannot reproduce this in cabal 2.3. Is this a 2.0.0.1 bug?
      • $HOME/.cabal/config has optimization: False.
      • Is this a regression? Oversight? It works with cabal install.

      • What I'm trying to do:

        • Build transitive dependencies with optimization disabled, for faster development.

      • My guess:

        • There seems to be a problem in how the new code path plumbs down arguments.

        • "Using internal setup method with build-type" always gets argument --enable-optimization.

          • That message is printed by ./Distribution/Client/SetupWrapper.hs:418 internalSetupMethod if cabal new-repl is run with --verbose.
          • Where does that --enable-optimization come from?
          • Why isn't the --disable-optimization passed down?

      • Related issues:

  • Cabal codebase

    • Seemingly minor codebase maintenance problems

      • Code duplication

        • CmdRepl.hs seems to be copied from CmdBuild.hs.

      • CmdBuild.hs imports ...Orchestration unqualified without explicit import list

  • Cabal description field pitfall

11Using Kubernetes

12Using Trello

  • How should we best use Trello?

    • A card gathers people to help each other accomplish a common goal.
    • A card disseminates information from the people who have it to the people who need it.

  • Traps

    • Because Trello cannot collaborative edit, we must not write too long before we save. Two people must not edit the card at the same time.

  • What do we use Trello for?

    • Prioritize what to do, starting with the highest velocity (value per effort)
    • Give enough information for others to help us

  • What Trello is not

    • A Trello column is not a mere to-do list.
    • Milestone lists are better than to-do lists. Be declarative, not imperative. Describe what you want, not how you want to do it.
    • If you can do it in a few hours, you don't need to make a Trello card for it.
    • If you don't need help with something, don't write a Trello card for it.

  • Our attitude towards Trello

    • We should consider the time spent on Trello as overhead. We should minimize the time we are editing, moving, arranging Trello cards.

      • Human nature trap: It's fun to look busy (over-organizing).

  • When not Slack?

    • If you need to remember it later, use Trello or Confluence, not Slack.
    • If a problem will take some time, don't waste your time typing on Slack. Put it in Trello, and assign it to the person who can fix it.

  • What is in a card?

    • Write a card with the intention of helping everyone else help you. The card must answer "What can I help? How can I help?".
    • A card contains information or will contain information.

    • A person should be in the card if:

      • he person needs that information, or
      • he can contribute that information (he may also contribute in a comment).

    • No other person should care about that card.

  • What is Trello?

    • To answer this, we answer these questions:

      • What does Trello make easy?
      • What does Trello make hard?
      • We seek an operational definition for Trello. We define something by what it do.

  • What should we group together? Why?

    • A board is a list of a column.
    • A column is a list of cards. Trello uses the term "list", but we use "column" to avoid confusion.
    • A card has zero or more members.
    • A card has zero or more labels.
    • A label is like an atom in propositional logic.
    • The filtering system is a limited form of propositional logic.
    • A card is a way for several people to share information.
    • If two people need the same information, they should subscribe to the same card.
    • If we need two people to finish (archive) a card, then both of those people should be members of that card.
    • What is a label? You give a label L to a card if and only if you often need to filter (select) all cards labeled L.
    • Why are two cards in the same list? Because they have something in common. They share something. An aspect of them is equal.
    • We group something in a list to minimize moving cards.
    • What is that aspect? What irreplaceable advantage does it give us?
    • Because they have the same members? (We can use filter for this.) Because they belong to the same team? Because they have the same due date? (We can use filter for this.) Because they are a part of the same user story? Navigating multiple boards is hard (big cognitive burden).
    • Label is for filtering.

  • What is easy to do in Trello?

    • Toggle any of the first 10 labels.
    • Show cards assigned to me (the user who is logged in).
    • Show cards by a conjunction or a disjunction of a Condition.
    • Condition is an element of Conditions.
    • Conditions is the union of all labels in a board and members of a board.
    • Click on the due date in the card description to mark it complete.
    • Read card comments in reverse chronological order (newer comments first).
    • Move a card to the bottom of an adjacent list.
    • See the number of checked items in a checklist in a card.
    • In the comment of a card, add another card.
  • Trello assumes that every checklist in the item has the same effort. Therefore we must make sure that every item in a checklist in the item has the same effort.

  • What is a bit hard to do (because it cannot be done by keyboard alone):

    • Adding a check list in a card
    • A card can have many checklists
    • Marking a check list in a card
    • Unmarking a check list in a card
    • Filtering makes navigating a large board possible.

  • What is hard to do in Trello:

    • Move from a board to another boards. (High cognitive load due to context switching.)
    • One person should not be in more than one board.
    • Move a card from a list to another list. It is easier to archive the list.
    • Unarchive a card

  • What is impossible (assuming no plugins):

    • See due dates in non-US format
    • Use keyboard to move card up or down in a list
    • Gantt chart
    • Calendars
    • Progress report
    • Collaboratively edit a card description. Trello will only show the last saved description. The previous descriptions are not lost, but hardly accessible. The old description can be accessed by Exporting a card as JSON.
    • In a card description, we can link to a board or a card, but not link to a list.

  • Member vs subscribe:

    • If and only if you are a member of a card, it will show in 'my card' filter (Press Q) The members of a card are the people assigned to a card We must tailor our workflow so that we use only the easy things.

  • Please read and memorize the Trello shortcut keys.

  • In search of a Trello architecture

    • Fixed Pipeline / Mini waterfalls

      • Column = team, Card = product
      • One column is assembly line.
      • Cards move from left to right.
      • Progress is sequential. Right column cannot start before left column finishes what it has to do with the card.
      • Ideal for the same process that is repeated very many times.
      • For example, in our case, the columns would be Requirement, UI, Database, Backend, Frontend, Done.

    • Column = a rather big to-do, card = breakdown of the column

      • For example, in our case, each column would be "As a (who), I (do what)".
      • A column is a work item and each card in the column is a breakdown of that item.
    • Column = epic / bigger user story, card = smaller user story

    • Column = milestone, card = breakdown of milestone

  • Trello list hierarchy (a Trello account is a list of lists of …)

    • A Trello account is a list of boards.
    • A board is a list of columns.
    • A column is a list of cards.
    • A card is a list of checklists.
    • A checklist is a list of checklist items.
    • Therefore we can use Trello for work breakdown structure of perhaps at most 4 levels deep.

13Using XML

13.1Finding the editor

List of editors:

  • I have used

    • Vim
    • IntelliJ IDEA

  • I don't know

Sources:

IntelliJ IDEA works fine, but I'd be happy if there is a lighter alternative.

My criteria:

  • open-source
  • schema-aware, supports XML Schema Definition (XSD)

  • autocompletion

    • 2018-08-22

Nice-to-have features:

  • Automate closing tag.
  • Editing an opening tag also edits the closing tag.
  • Editing a closing tag also edits the opening tag.
  • Any way to avoid memorizing and typing xmlns:xsi.

Vim can semi-automate XML closing tag using Ctrl+P.

13.2Using schemas

In principle, from a type defined in an XSD file, we can generate a Haskell module and a Java class.

13.3Toward markup language agnosticity

What is common between XML, JSON, and YAML?

Can we map (interconvert) between XML and YAML? Approximate data model:

Can we use XML Schema to validate a JSON/YAML document?

We can map a subset of XML to YAML using YAXML, the (draft) XML Binding for YAML.

14Using PostgreSQL

14.1Traps, pitfalls, gotchas

14.1.1PostgreSQL ALTER TABLE DROP COLUMN doesn't actually remove the column from disk

From PostgreSQL Tables Can Only Have 1600 Columns, Ever. | Nerderati.

15Invert screen color

https://askubuntu.com/questions/181419/how-to-reverse-colors-for-the-current-window-in-gnome-shell xcalib -invert -alter Keyboards Bind it to Super+I Bind gnome-terminal to Super+T

16<2019-05-19> Time to buy a new computer

My old computer's left DIMM socket is broken, and 4 GB RAM is somewhat restrictive.

I am looking for a laptop with these features:

  • zero-maintenance
    • no dust; fanless
  • serviceable
    • replaceable battery
    • replaceable disk
  • at least 8 GB RAM
  • long durable battery life with adjustable full charge
  • SSD, at least 64 GB; I will set up encrypted LVM and download development tools
  • service center
  • warranty
  • quick service (solve in hours, total replacement if necessary) no matter what the problem is
  • fits in my bag, easy to carry, but not too small as to be unreadable
    • lightweight
    • dimensions?
  • no fatal negative online reviews
  • not too new, lest it be defective

Indocomtech 2019 will be held in 30 October – 3 November.1

17<2019-08-20> Howto: Send gmail as namesilo email

namesilo: "Please note that we do not offer a service for sending email."2 Follow these instructions to set up gmail3.

There are two SMTP servers. Receiving and sending use different SMTP servers. Receiving email uses Emailowl SMTP, but sending email uses Gmail SMTP. Set up Gmail "app password".

  1. https://indocomtech.net/dates-hours/↩

  2. https://www.namesilo.com/Support/What-Email-Service-is-right-for-me%3F↩

  3. https://www.namesilo.com/Support/Gmail-Instructions-for-Reply~to-Using-Custom-Domain↩