• 0 Posts
  • 112 Comments
Joined 1 year ago
cake
Cake day: June 5th, 2023

help-circle
rss


  • I switched nearly two decades ago after I used a freeware network monitor on Windows and realized that it was making dozens of silent TCP connections online. Some were to Microsoft, while others were to unknown third parties. Just imagine your personal machine doing this!

    Linux is actually easy to use these days. Installation is often easier than windows and hardware just works most of the time. Despite that, people have a habit of exaggerating the difficulties in using Linux or BSD. They very often feel like excuses to avoid checking it out.








  • My sincere belief is that the difficulty in self hosting is due to the lack of priority, investment and development, due to the perverse incentives of the SaaS model. I don’t think it’s a technical problem that cannot be resolved with sufficient work. There are PoCs that prove that it can be made as simple as desktops and mobile phones.


  • That page pitches Nomad as a direct and better competitor to K8s. Both are considered as container orchestration platforms, though nomad can orchestrate other types of jobs as well.

    When it comes to scalability, the anecdotes I’ve heard says that Nomad is better. Even the page you provided says the same. (I did try Nomad. But didn’t scale it enough to test this).

    The only real issue that I faced with Nomad in comparison to K8s is running certain infrastructure loads like CNI and CSI plugins (like longhorn and mayastor). They don’t just talk to K8s through the standard interfaces (which Nomad also has), they often integrate deep into K8s using operators and CRDs. Nomad doesn’t have the provisions to support such nonstandard deep integrations.


  • I have to disagree with both those assertions.

    If a software is easy to self host, then there is no need to make it harder to deploy as SaaS. The latter will be irrelevant for most people.

    And the problem of self hosting isn’t a circular problem as you project it to be. There are architectural changes that can make it positively easier to self host without exposing the sysadmin to needless complexity. The example I quoted before - sandstorm - was a step in this direction. Deploying and administering applications on sandstorm would have been as easy as deploying one on desktop (including cross app integrations). The change needed was to modify the app to work with the sandstorm platform. Unfortunately, the platform didn’t gain the momentum needed to ensure that all available apps would be ported. But it shows that the concept is viable.





  • The fundamental problem I see here is the cloud. We were supposed to have easy self hosted applications and data on cheap always-online hardware. Instead, companies promoted cloud services with the intention of rent seeking through subscriptions.

    If you look at the software that went from open source to source available, you’ll notice that almost all of them are cloud applications. Why? The companies that created them were hoping to make money through the same subscription model. But then, big cloud players like AWS just outcompeted them using their own software.

    Would this have happened if the FOSS ecosystem neglected the cloud hype and gravitated towards self hosting? Perhaps. But not as badly. We still haven’t seen enough progress towards self hosting. It’s still very hard for regular folks. Genuine efforts like sandstorm didn’t find enough momentum. I hope this changes at least now.