1,085 karma · joined December 9, 2014
Software Architect, Python/Django enthusiast.
Remote: Yes
Willing to relocate: Depends
Technologies: Python, Django, Tailwind, C#/WPF, C++/Qt, PostgreSQL, AWS (EC2, DynamoDB, ElasiCache, RDS, Lambda), Terraform (AWS, Azure), Ansible, Linux, Windows, Git, TailwindCSS, reverse engineering formats and protocols, cryptocurrencies.
Résumé/CV: https://linkedin.com/in/adontz
Email: adontz@gmail.com
I am an Engineer with experience in backend software development, desktop software development, embedded development and electronics, infrastructure management. In recent years I was head of devops department in a bank and CTO is a startup. Seeking for something less stressful. Love open source. I like to know things.
Remote: Yes
Willing to relocate: Depends
Technologies: Python, Django, Tailwind, C#/WPF, C++/Qt, PostgreSQL, AWS (EC2, DynamoDB, ElasiCache, RDS, Lambda), Terraform (AWS, Azure), Ansible, Linux, Windows, Git, TailwindCSS, reverse engineering formats and protocols, cryptocurrencies.
Résumé/CV: https://linkedin.com/in/adontz
Email: adontz@gmail.com
I am an Engineer with experience in backend software development, desktop software development, embedded development and electronics, infrastructure management. In recent years I was head of devops department in a bank and CTO is a startup. Seeking for something less stressful. Love open source. I like to know things.
Is there a non public version with very reliable arm/mil? Is there a version which can reliably distinguish T-80 with and without Z?
Telegram is used by drug dealers in my country too. All types of illegal content. Honestly Telegram is new darknet, but not even that dark. Streets are full of QR codes linking to Telegram channels,
OTOH Telegram is a really good technology. Very lightweight client. Whatsapp sucks compared to Telegram. Very hard to block protocol, Russian authorities basically failed to block Telegram. Very easy to write bots. Telegram bots are very powerful, way beyond simple chat replies. One can build UI with buttons and dropdowns inside Telegram.
Also, in not so free countries Telegram is the Internet basically.
WHERE CURRENT_TIMESTAMP BETWEEN start AND end is much better than WHERE deleted_at IS NOT NULL
Also, if we are talking about real use case of audit, not a simplified artificial one, my real tables were more like
changeset(id INT, timestamp DATETIME, user ^USER, host TEXT, ip_address TEXT, <other security data>)
datatable1(id INT, created_at ^changeset, updated_at ^changeset, deleted_at ^changeset NULL, <other fields>)
datatable2(id INT, created_at ^changeset, updated_at ^changeset, deleted_at ^changeset NULL, <other fields>)
datatable3(id INT, created_at ^changeset, updated_at ^changeset, deleted_at ^changeset NULL, <other fields>)
Are not we talking about PostgreSQL?
Clones/double use/double spend must be caught on reader/server anyway. One can pass a card to another person, and you do not want two people to enter building with the same card.
I implemented double spend protection by introducing a simple operation counter. If the sequence of operation IDs is not continuous, card is blocked. Clones were added to block list within minutes. It was good enough for the use case. Again, MiFARE is very cheap, so tradeoffs are expected.
1. There may be audit policies, compliance policies, retention policies in place. You may HAVE to keep data which is deleted from user's perspective.
2. People make mistakes. While I agree that restoring a long ago soft-deleted object may be very tricky because of referential integrity. It's easy to restore a recently (hours, days) soft-deleted object. Not only easy, it's much faster than restoring a point-in-time backup of an entire database. Also restored version is guaranteed to be the latest one, unlike in backup.
3. PostgreSQL supports partial indexes and partial unique constraints. https://www.postgresql.org/docs/current/indexes-partial.html
Other databases support similar features too. That allows excluding soft-deleted records from indexes, makes indexes smaller. A good query does not scan a table anyway. If one needs to read less pages from a table, PostgreSQL has CLUSTER command. If there are too many soft-deleted tuples, maybe it's time to think about an archive table anyway.
Unique constraints may ignore soft-deleted objects, so only active objects need to be unique. That is very useful if you need to support idempotent calls. Additionally that may actually be a security feature too. If I create a new user with the same username as an old deleted user, I know that username is unique, but also, that I will not accidentally get permissions of the old user assigned.
4. The worst is tooling. Forgetting "WHERE deleted_at IS NOT NULL" or "WHERE NOT is_deleted" is a very real problem. I would rather advise against soft deleted objects because of that problem alone.
On the other hand, if you use a decent ORM, (and why would not you?) like SQLAlchemy, Django ORM or EntityFramework/Linq (sorry for guys who are not Python/.Net, I do not have a good example for you), then it's quite easy to create a default model query which is not empty. But honestly, if your system is complex enough and has any custom row level security, if it is multi-tenant, if not all users see everything, you already have quite similar problem, you already have to start with non-empty query which already has WHERE and possibly even JOIN clauses. Without a good ORM it will suck anyway.
I am huge fan of having a EC2 machine in AWS and using VSCode remote SSH or anything similar for development. There are so many differences between cloud and local, that you either embrace it (develop in EC2) or try to wrap it into one more layer or abstraction (hello Docker) and still fail.
QML is nice, animations were much smoother than I expected.
https://www.statista.com/statistics/263560/net-revenue-of-in...
- async/await runs in context of one thread, so there is no need for locks or synchronization. Unless one runs async/await in multiple threads to actually utilize CPU cores, then locks and synchronization are necessary again. This complexity may be hidden in some external code. For example instead of synchronizing access to a single database connection it is much easier to open one database connection per async task. However such approach may affect performance, especially with sqlite and postgres.
- error propagation in async/await is not obvious. Especially when one tries to group up async tasks. Happy eyeballs are a classic example.
- since network I/O was mentioned, backpressure should also be mentioned. CPython implementation of async/await notoriously lacks network backpressure causing some problems.