Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
elithrar
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
21 ms
·
121.
▲
by
elithrar
7y ago
This is an extremely unconstructive comment. It's easy to be critical with 20/20 hindsight. More constructively: what would you have picked back in _2011_ when Cloudflare was getting off the ground? Ideally, it needs to have a mem
122.
▲
by
elithrar
7y ago
Further, sometimes users aren't truly aware of the implications of what making a bucket "public" means. Heck, maybe the original use-case wasn't too concerning, but it changed over time. The folks writing/reading fr
123.
▲
by
elithrar
7y ago
> First, OpenConnect is just a CDN run by Netflix, they could call it bunny protocol it’s just a name. But, they don’t have unlimited boxes at every ISP, in almost every case you can get P2P connections between specific users with lower
124.
▲
by
elithrar
7y ago
The parent conflates copyrights with trademarks.
125.
▲
by
elithrar
7y ago
You’ve hit the nail on the head. It’s a “technically wrong” term that conveys enough meaning: unlit fibre you buy and light up yourself (usually).
126.
▲
by
elithrar
7y ago
Neither is “great” on a GPU: • Image optimization is still mostly software: JPEG, PNG, WebP. GPUs do not bring much to the table here at all, and consume a lot of power, which is likely to upset the hosting ISP of that cache node. • Video e
127.
▲
by
elithrar
7y ago
As a Googler: take VM rightsizing recommendations - "save $X because you're underutilizing this machine shape" - and extend them to encompass this by including VM-family swaps based on underlying VM metrics? :)
128.
▲
by
elithrar
7y ago
That still breaks the experience. Now you have tampered content AND broken clients. TLS does wonders for preventing transparent proxies, DPI & ISPs from modifying or tracking what you see.
129.
▲
by
elithrar
7y ago
Disagree. - TLS is still important to stop tampering of video content or images, as well as user privacy over what content was specifically viewed. - Some ISPs have (and still do) intercept plaintext video content -> transcode to a much
130.
▲
by
elithrar
7y ago
Correct. Results are cached for the same program “ID”.
131.
▲
by
elithrar
7y ago
This is categorically untrue. They are not a secret due to the approvals required at a local level and by the FCC. Many carriers lease space on the same towers, or have agreements to lease space from others in certain markets. You almost ne
132.
▲
by
elithrar
7y ago
I’m not sure why the other post was marked dead, but AutoML models can be exported - you can export them to TFLite format[0] and then run them on edge devices, such as a SparkFun Edge[1] or a Coral SoC / board. [0]: https://
133.
▲
by
elithrar
7y ago
Some customers also believe they want this: CDNs are often about both performance (content closer to user) && reduced origin load. Caching a (too!) broad set of error codes “achieves this”, with the caveat of a poisoned cache.
134.
▲
by
elithrar
7y ago
> Reading what cloudflare sells, people who have a domains on Cloudflare and buy Pro, Business and Enterprise plan have access to DNS analytics that displays DNS queries from Cloudflares DNS servers, including geographic information. Tho
135.
▲
Gitlab Customer Acceptance Policy
(gitlab.com)
5 points
by
elithrar
7y ago
|
0 comments
136.
▲
Twitch’s New Brand and Logo
(blog.twitch.tv)
10 points
by
elithrar
7y ago
|
2 comments
137.
▲
by
elithrar
7y ago
I'm consistently amazed at how many people [on HN...] over-estimate the size/scale of Cloudflare. Comcast is a huge, multi-billion $ operation with global contracts and broadcasting capabilities, and > 184,000 employees globall
138.
▲
by
elithrar
7y ago
That feels like a headline feature - cards not being tied to specific humans at the org is great for the (SaaS/IaaS/Cloud) reasons above. The 1:1 person:credit-card mapping for those without invoice billing will continue to cause
139.
▲
by
elithrar
7y ago
IIRC that's never been a public API; more of an artifact for troubleshooting/support. Relying on it / parsing it directly was never recommended when I was there.
140.
▲
by
elithrar
7y ago
Paragraphs 3, 4 & 5 of TFA speak directly to the benefits of a module proxy and mirror w.r.t. module availability & performance.
141.
▲
by
elithrar
7y ago
The parent is inferring that Lukas owns a non-insignificant stake of First Solar , a quasi-competitor of Solar City/Telsa, not Walmart themselves.
142.
▲
by
elithrar
7y ago
> Would be interesting to get a pie chart for costs between US and China like land acquisition costs, labor, raw materials, management, etc. Safety, incidents/deaths, etc... (Although I believe that the costs for large infra project
143.
▲
by
elithrar
7y ago
I think the part here is that there aren’t tools “like” gofmt - there is only gofmt. Other tools wrap it, but there’s no choice (and bikeshedding) to make.
144.
▲
by
elithrar
7y ago
Also: if they switch to another app, platforms like iOS will pause/suspend (within a few seconds) the upload before it likely completes.
145.
▲
by
elithrar
7y ago
I want to be clear that my question relates to artifacts you produce on company-owned resources - not just in your own time on your own machine. Legal precedence around IP ownership - when you've used company-owned machines - is far
146.
▲
by
elithrar
7y ago
> while still being officially allowed to take the devices home with us, use them privately and so on. Who owns the rights to IP developed on these company-owned laptops? One of the biggest problems with this kind of ‘unspoken flexibilit
147.
▲
by
elithrar
8y ago
> But it seems it would be rather cheap to set up an infrastructure made principally of steel towers and (thin) cables, and there shouldn't be many technological challenges. Just huge logistical challenges with right-of-way, constru
148.
▲
by
elithrar
8y ago
You can’t always predict how you will react under pressure. If you have an equipment failure or injure yourself in a way that you have not prepared for, well...
149.
▲
by
elithrar
8y ago
The ask here was inference, not training.
150.
▲
by
elithrar
8y ago
The USB accelerator is designed for that - so you can target models for the a (small scale) TPU and then scale up on the cloud. This would specifically let you make sure that the TensorFlow ops your algorithms use are supported on a TPU.
More ›