I’d think if CF were interested in branching out (no pun intended) into auxiliary areas, they would be looking into perhaps offering some Netlify-like services to complete the circle (or even acquiring Netlify, which I believe would be a perfect fit for CF).
After clicking through, this was pretty obvious. I think there are just a lot of fired up hackers out there wanting to see WW3 between Github, Gitlab and the rest of the world.
I know this would consume a lot of bandwidth, but it would always be nice to see more official package repo mirroring, like in front of a github repo / npm / rubygems, PyPI etc.
I feel like right now we are in a fairly centralized state regarding where packages come from and outages and other issues can really impact the CI/CD and development of different services.
I do have a script that finds all mentions of Cloudflare in stories and comments and emails. Latency between a mention and me getting mail is about a minute.
well you could help out gitea and make a hosted version on gitlab.
there is a ton of stuff which cloudflare could do, like creating a object storage git backend, etc..
Cloudflare folks. Please don’t do this. I don’t know if Mr.Cumming is serious.
As a user of CF, I hope you folks go a lot deeper than broader. Example: Better cache management – compete with Fastly, my tickets on load balancer bugs were open for almost 6 months, etc.
Add to that list: hardware-based 2FA. Based on past tweets from jgrahamc, I believe CloudFlare requires all their employees to use hardware U2F tokens for critical things like Google services and has for awhile, yet they still don't support those on their own dashboard.
To tag onto what JGC said, you have complete programmatic access to the Cloudflare cache inside a Worker [1] using the Cache API [2]. If there's anything VCL can do which Worker's can't we would love to hear about it. We also support Terraform for all of our APIs if your goal is to have a config language to configure Cloudflare settings.
You can use Workers to write real code on our edge. VCL is ok but it’s not coding. People are writing real programs on our edge not just configuring it. But you can also use Workers for that.
The amount of business Cloudflare employees handle direct over Twitter, and the amount of responsiveness from their higher-ups on it is... kinda incredible. I imagine it's only practical because Cloudflare doesn't have a lot of public brand awareness beyond tech folks.
If only you could say the same about potential employees. Your recruiters seem to have a penchant for completely ghosting candidates, even when several interviews deep into the interview process. This has happened to me twice now, with two different recruiters, even after being told by the interviewers that I was well qualified and surely would be moving to the next step. And after glancing at Glassdoor reviews, this doesn't seem to be uncommon whatsoever.
I'd like Cloudflare to not do a Google and let a thousand flowers bloom only to let most of them wither. They seem to be taking on a lot just with their new public VPN service.
One feature I would love is a global code search that has the quality of the repo (as measured by forks/stars/pagerank...) as a part of relevance ranking. A use case is simple: I have some library and I am not exactly sure how to use it, so I go into global code search and try to find usages of that method. First thing that comes would be some relevant code in a big context of a well engineered repository.
Precisely. For my organization, I'm looking for a way to quantitatively measure and compare quality/security postures of repos/packages when importing.
Sounds tough. What are your ideas to measure so far?
For the security angle it seems the first naive thing would be to count prior known vulnerabilities, but then, the projects that do absolute worst at that are not going to have discovered their security bugs let alone document them well.
Decentralized SCM features for issues, CI, etc. GitLab is great but it is still mostly centralized. Creating a modern connector for IPFS would be awesome.
cloudflare running SMTP relays for your own mail system, in both directions, doing spam filtering, etc, could be valuable. argo, edge workers, etc could all be useful in an email environment.
most everyone is already convinced you can't run a mail server yourself, so perhaps the haters wouldn't think that cloudflare is ruining anything by getting involved in mail?
The most useful thing Cloudflare could do is allow us to disable the `__cfduid` cookie. That's one of the big reasons I stopped using Cloudflare on a public-facing API/CDN.
Just curious, what is your concern? Is it privacy related? Or does it inhibit consumption of APIs you tried to host with them? I don’t like the header either and I have a public API with Cloudflare but it hasn’t caused any technical issues as far as I know.
The only non-core-business work that I really want from Cloudflare is to do something with LuaJIT stewardship, because I thought they were supposed to be taking the helm and the official GitHub repo is basically completely dead.
More seriously. I’d be curious to hear about what you’d build on our network. We are all over the world and it would be fun to hear about companies other than GitHub that you think we should disrupt.
I'm a big supporter of Cloudflare and have been using it for almost 8 years. I personally don't think doing something like that is what Cloudflare needs.
What Cloudflare needs is further customization, especially in regards to caching. We actually had to migrate a certain part of our infrastructure to Fastly due to the lack of caching customization/rules.
I'd like to see:
- Custom caching rules similar to the new firewall rules
- Finer granularity of the cache expiry (I'm aware Enterprise has the ability to cache for 30 seconds, but we don't want to upgrade to Enterprise just for that one thing).
- Cache hit rate analytics grouped by path/domain/etc
I've submitted suggestions to some people but I believe with the workers it should be possible to create your own custom pull/push CDN because of the granularity controls they have added. It used to not be possible to interact with the caching layer but since they have added controls for that (about a year or so ago)
Maybe this is already in the works, since the enterprise plan has a logging feature, but an integrated logging solution would be nice for the other plans. Cloudwatch isn’t perfect, but logging from a lambda worker to cloudwatch is very easy compared to, say, having to setup/manage an ELK stack for this purpose on Cloudflare.
I’d like to be able to have more than one unique Cloudflare worker per account. And be able to assign specific workers to specific routes. This feels like an odd limitation and forces you to make huge workers code with path case statements to separate the request actions out. Pretty soon a single worker starts to feel like a large app instead of a simple function that does mostly one thing.
The ability to have more than one worker or at least a “dev” worker is really showing to be important to me. Especially the more logic that is moved to the edge, the more complex it gets, and the more testing/review might be needed so you need some way to stage changes before promoting to production.
Overall though I really like Cloudflare workers. I’ve moved a project that was built on AWS and it’s mostly all an improvement on CF, especially performance which was very noticeable. I run one public API entirely on workers and I use workers on the public facing site to scrub the path users send us and 302 them to the right place, instead of having to do that in the app. I also remove most query params inside the worker to improve caching performance. I should have a use case for the KV system soon.
Forgive my ignorance but does CloudFlare have any technical or infrastructural advantage to address problems like the great firewall? Or provide options for people when their nation states lock down the internet? Access to a free and open internet seems like a worthy moonshot to pursue.
I know everyone love to hate on the industry but I think with such firepower, I'd go for a "takeover" of the 3rd parties (read middle men) in Online Advertising.
You probably have decent capabilities to detect fraud while building a privacy minded tracking system where the user can manage its personas.
While the ITP (Safari's -and maybe Chrome soon-- attempt to regulate tracking) initiative is on paper a good idea, I'm not convinced it's going to deliver.
Barging in with an OpenRTB v4 (current version is V3) that delivers on that - and maybe cut the middle men, that could be neat for every one : publishers, consummers and advertisers.
Do you plan on offering services below the HTTPs/QUIC layer? I am thinking raw UDP/TCP. That way apps that are not HTTP protocol driven can piggy back on tunneling, security and possibly short cached CDN.
Are there any updates to the Cloudflare Registrar?
Would be awesome to have marketplace or master/sub accounts. That is, launch a saas say a landing page builder and then allow my customers to also purchase domains from my saas (under my brand account) and have them automagikally configured. Instead of having the user to use namecheap/godaddy and mess around with their crappy control panel. Maybe have an opportunity to raise the price by a $1 [to the end user] and have revenue stream?
We do have a product called SSL for SaaS which handles most of the components of that flow like DNS and TLS certificates. The actual domain registration will take a bit longer to integrate, but I'm guessing you could make it work using our API.
2) A page rule for “don’t send origin cookies” or something (useful for not passing through those headers from the origin for assets/etc)
3) getting the peering connections better in AU so the network handles traffic like normal instead of only $200/mo plans getting priority/local AU routing
If Cloudflare wants to branch out, here're an idea.
Global Memcached or Redis. Support caches near the edges. SPA or apps running on a client's browser/machine can use the one nearby. Support pre-populating/warm-up the caches in batch.
How about Github without a big corporation? Not controlled by Cloudflare, ddecentralized and uncensorable. Like bittorrent, but for git repos, issues, etc.
So... git? It's already designed to be decentralized.
Hosting a public read-only repo would be as easy as standing up NGINX and pointing it at your repo folder, afaik.
At my previous job, we all had individual dev machines that we each had accounts on. So, we each had a copy of our repos that we could push to/work from. I wrote a little git alias that would add remotes for each others' machines so we could pull directly from each other (though we couldn't push).
Hell, even a "private" git repo wouldn't be too hard as long as you don't care about per-project permissions. Just make a user (e.g. 'git') that has read/write access to the folder and add some SSH keys to authorized_keys. Shazam, you have a private, self-hosted git repo at ssh://git@yoursite.com:path/to/repo.git.
Bare git only stores code, not issues, tickets, wiki, and all the stuff that goes beyond source files. There are a myriad of open source solutions for this, but there is no standard format that allows easy migration and inter-linking.
Well for wiki, usually a branch containing markdown will transfer well across forges. Issues and PRs are the real issue. I’ve been hoping git appraise would get supported by the forges for distributed PRs, but it doesn’t look like that’ll happen.
Very interesting that gitlab is standardizing on the de facto standard of patch requests! How ironic it is that the existing git-request-pull isn't used more than needed thouh
would be easy enough for a branch of "issues" containing a number, index.json and ####.md where the number is the initial issue and comments, with some front-matter to associate user details.
To be clear, there would be directories, each directory being the next incremental number representing the issue number, and the directory containing numbered markdown documents with yml front matter starting at 0 being the root of the issue thread.
I'm kind of surprised it isn't already like this, or similar.
Huh? So we are going to jump on smaller corporations, make them bigger and leave them for a smaller one? I believe the keyword in his idea was decentralized.
But I’m happy to hear all the suggestions.