5 ms·

Github relies on a distributed storage architecture (local git repositories on developers' machines) and in theory anyone can take a project from github and dup

Github relies on a distributed storage architecture (local git repositories on developers' machines) and in theory anyone can take a project from github and duplicate it on gitlab etc. In reality nobody bothers and a project hosted on github will remain exclusively hosted on github and nowhere else, and likewise with other git hosting services. For the most part nobody cares if the data is hosted on a distributed system or a centralized one, because the overwhelming majority of users will rely on the front end. Changing RPC urls is not as easy as you might think, especially for systems that are widely deployed and have heterogenous clients (which in theory would be the case if Web3 ever took off, which I personally doubt).

sfblah 5y ago
And, as the article suggests, if there is some new feature that Github can enable (integration of git commits with an issue tracker or CI/CD integration come to mind), that will happen in a vendor-specific way on Github, not in the Git protocol. So, then you immediately move back to the world of platforms.
pshc 5y ago
Few people bother because every git commit is cryptographically signed and every git repo is inherently replicated. It doesn’t matter if you use a centralized service or not as long as you can rely on SHA1 (and sha256 is coming…) Git is almost the ur-blockchain in this respect, hardly an argument for centralization. Also, fuck Microsoft.
anderspitman 5y ago
I keep my projects on GitHub for discoverability and the reputation provided by stars, whatever that's worth. So essentially network effects. The reality is if I'm looking for a library to solve a problem, I'm much more likely to use one from GH with 1000 stars than a random self-hosted GitLab with 50. I would like to not feel that way, but I suspect many others do as well. It would be nice if we at least had a decentralized reaction/reputation system. Is there an analog to this with the services Moxie talked about? Sincere question, I'm not familiar with the ecosystem at all.
- clippablematt 5y ago
  The closest attempt I can think of is status wallets token ranking for dapps. You could burn your tokens to say if you liked something in their listings and that would rank it for others. The issue they hit is when the lists got popular (in like 2018-19) vc funded projects just bought up the supplies of tokens and burnt them to get their project rated higher. So basically Sybil attack and they became unreliable. It’s the same problem across all decentralised protocols, if it’s cheap to say something you get spam(see email) but introducing costs can just skew it to those who can afford to spam instead (essentially those with an advertising budget). So there’s been a lot of research on proof of personhood (BrightID/ideas/proofofhumanity) to add Sybil resistance mechanisms so we can do 1p1v across the network. They’re working ok, but the next big step is adding zkproofs so we can anonymise the voting (which is needed to prevent collusion) which clrfund and sismo are working on. Kleros have an interesting curated register protocol, which seems to work on small scales. Some groups are using it to token rank guy issues to prioritise work and get feedback. The status blog has some interesting writing around these ideas over the past few years https://status.im/research/ https://status.im/research/
strken 5y ago
People do bother. I don't have specific examples off the top of my head, but I've occasionally run into an read-only GitHub repo that's been moved to Gitea or GitLab, or even BitBucket. More broadly speaking, it's important that you can migrate, even if you don't actually do it, because users who can easily churn give the developers an incentive to keep the UX solid. If you can just leave GitHub at any time, then they're less likely to add gigantic banner ads to every page, or bundle "third party offers" into installers - they know what happened to SourceForge, after all.
- spullara 5y ago
  That is like saying that some people bother to host their own email.
  - strken 5y ago
    Not really. The barrier to hosting your own email is that you'll spend a day configuring everything, and a year later, the big providers will slightly change a spam detection algorithm, your mail won't be delivered, you won't know, and there will be bad consequences for you. The barrier to changing your git origin is spending five minutes setting up an account and repo somewhere else. Everything will work absolutely fine, you'll still have all your git history, you'll just be slightly less discoverable and some potential contributors might not want to create an account.

sfblah 5y ago

And, as the article suggests, if there is some new feature that Github can enable (integration of git commits with an issue tracker or CI/CD integration come to mind), that will happen in a vendor-specific way on Github, not in the Git protocol. So, then you immediately move back to the world of platforms.

pshc 5y ago

Few people bother because every git commit is cryptographically signed and every git repo is inherently replicated. It doesn’t matter if you use a centralized service or not as long as you can rely on SHA1 (and sha256 is coming…) Git is almost the ur-blockchain in this respect, hardly an argument for centralization. Also, fuck Microsoft.

anderspitman 5y ago

I keep my projects on GitHub for discoverability and the reputation provided by stars, whatever that's worth. So essentially network effects. The reality is if I'm looking for a library to solve a problem, I'm much more likely to use one from GH with 1000 stars than a random self-hosted GitLab with 50. I would like to not feel that way, but I suspect many others do as well. It would be nice if we at least had a decentralized reaction/reputation system. Is there an analog to this with the services Moxie talked about? Sincere question, I'm not familiar with the ecosystem at all.

clippablematt 5y ago
The closest attempt I can think of is status wallets token ranking for dapps. You could burn your tokens to say if you liked something in their listings and that would rank it for others. The issue they hit is when the lists got popular (in like 2018-19) vc funded projects just bought up the supplies of tokens and burnt them to get their project rated higher. So basically Sybil attack and they became unreliable. It’s the same problem across all decentralised protocols, if it’s cheap to say something you get spam(see email) but introducing costs can just skew it to those who can afford to spam instead (essentially those with an advertising budget). So there’s been a lot of research on proof of personhood (BrightID/ideas/proofofhumanity) to add Sybil resistance mechanisms so we can do 1p1v across the network. They’re working ok, but the next big step is adding zkproofs so we can anonymise the voting (which is needed to prevent collusion) which clrfund and sismo are working on. Kleros have an interesting curated register protocol, which seems to work on small scales. Some groups are using it to token rank guy issues to prioritise work and get feedback. The status blog has some interesting writing around these ideas over the past few years https://status.im/research/ https://status.im/research/

strken 5y ago

People do bother. I don't have specific examples off the top of my head, but I've occasionally run into an read-only GitHub repo that's been moved to Gitea or GitLab, or even BitBucket. More broadly speaking, it's important that you can migrate, even if you don't actually do it, because users who can easily churn give the developers an incentive to keep the UX solid. If you can just leave GitHub at any time, then they're less likely to add gigantic banner ads to every page, or bundle "third party offers" into installers - they know what happened to SourceForge, after all.

spullara 5y ago
That is like saying that some people bother to host their own email.
- strken 5y ago
  Not really. The barrier to hosting your own email is that you'll spend a day configuring everything, and a year later, the big providers will slightly change a spam detection algorithm, your mail won't be delivered, you won't know, and there will be bad consequences for you. The barrier to changing your git origin is spending five minutes setting up an account and repo somewhere else. Everything will work absolutely fine, you'll still have all your git history, you'll just be slightly less discoverable and some potential contributors might not want to create an account.