* Posts by _andrew

118 publicly visible posts • joined 7 Dec 2019

Page:

SiFive inches closer to offering a true RISC-V PC: Latest five-core dev board includes PCIe, SSD interfaces

_andrew

Strip off the decoder...

Not especially likely, as that would expose micro-architectural details that you probably want to be able to vary from model to model, such as number and structure of pipelines and size of re-order/re-name buffers.

Has been done before though, twice: Transmeta did essentially exactly that, replacing the x86 decode logic with a software dynamic compilation system that targeted an in-order VLIW processor to do the work. Nvidia's "Denver" cores and follow-up (as seen in the Nexus-9 tablet and several of the car-AI modules) do a very similar thing but for an Arm source-instruction-set.

Both work nicely on loopy, numerical code, but quite poorly on large, non-loopy code like user interfaces, database engines and operating systems.

Interestingly, Dave Ditzel was involved in both of those designs, and is now founder of Esperanto, a RISC-V company.

Is Google fudging search rankings to benefit pages that embed YouTube vids? Or is this just another ‘bug’?

_andrew

Re: iframe and probably object/embed too

I'm in favour of anything that down-rates pages that have different-size advertisements that make the text that I'm trying to read jump up and down as they change and force page reformats. Not that I use search or ratings to get to most of the web, but if big-G can use its influence on the ones that do, then I'm for it.

Google screwed rivals to protect monopoly, says Uncle Sam in antitrust lawsuit: We go inside the Sherman parked on a Silicon Valley lawn

_andrew
Coat

The default search on the default browser on the default operating system on every PC...

... is not Google. Yes they have the mobile market mostly stitched up, but that happened long after they were established as far-and-away the best search option.

What do you imagine will happen if Apple are forced to stop accepting Google's placement coin for search defaults? I'd expect that the _only_ effect will be that apple has to raise that extra revenue elsewhere, by putting up their prices: almost everyone will choose Google themselves, given the choice, because the alternatives are worse.

Worst outcome, I expect, will be the death of Firefox, which lives entirely on the income of that default search placement.

I remember the days of Yahoo! curated links and AltaVista. I'll choose Google any day.

After ten years, the Google vs Oracle API copyright mega-battle finally hit the Supreme Court – and we listened in

_andrew

Re: Declarations and implementations

Java doesn't have declarations, only implementations. The things that show up in javadocs that look like declarations are extracted from the .class files automatically. There _is_ no source file to claim copyright on in this case. The claim is over the structure and names of the base classes themselves, irrespective of their origin.

And to comment on a different topic up-stream: the fact that no-one has filed suit against GNU classpath is almost certainly that no-one cares, because it is costing no-one any assumed business. Or perhaps they will, once this case goes Oracle's way.

_andrew

Re: The case is about the APIs themselves

In Java there isn't any "API source code", as such: there is only implementation and the resulting .class files, and (pertinently here) the javadoc and interface description that can be mechanically extracted from the .class files. I'm pretty sure that that's what they're arguing about: the list of classes, methods and function names from the Java base library set. Names, arguments, structure. There isn't a source file with copyright notices on it, like a C header file. To be compatible, even a clean-room implementation would have to give you the same result.

_andrew

Re: The devel is in the details - Part II

As near as I can tell, this case does not turn on whether or not the APIs in question were a clean-room implementation or not. The case is about the APIs themselves, absent any implementation at all.

_andrew

Re: The devel is in the details

I think that you'll find that the border between "language" and API is fuzzier than you are making out. Not all programming interfaces exist just as function prototypes in a C-like syntax: some syntax is api. Every word in forth, smalltalk and lisp/scheme "languages" is a function API. Whether access to an API is by changing an instruction pointer value or a dialog in a particular serialised protocol (say, HTTP, or CORBA) is unlikely to make any significant difference to the principle at stake here. Sure, Matlab and SQL are languages, but they're also the APIs to a database and matrix engine, and within them there are both procedures and functions that have specific meanings. Now SQL has been standardized, but it originally belonged to IBM, and not all standards (even ISO standards) have free terms, let alone RAND, especially where patents are involved.

That OpenGL was open-sourced by SGI themselves may or may not be pertinent. Java was open-sourced by its creator too. Clearly the details do matter, but I think that reasonable people can disagree about the significance and meaning of specific details.

I'm not saying that there are impending court cases in any of these examples. Just pointing out that the activity in question is, on one form or another, common practice and has long historical record.

If we're really, really lucky, the ruling on the current case will be narrow enough that we (other than google and oracle specifically) can go about our business as before.

_andrew

Re: If Oracle wins, you lose?

Off the top of my head, and without checking any of the details and original license terms:

Any non-SGI implementation of OpenGL, probably including both WebGL and the Mesa library.

Octave and/or whatever common library APIs they reproduced from Matlab.

Win32 and DirectX reimplementations in Wine and Crossover and ReactOS.

Probably every SQL database implementation, including Oracle's...

Every VT100 (or derivative) compatible terminal, including xterm.

The list is not small, when you think about it a bit.

US senators propose yet another problematic Section 230 shakeup: As long as someone says it on the web, you can't hide it away

_andrew

There was editorial before socials

It's never that simple though.

You're always going to have to support take-down orders to comply with little issues like the law of various jurisdictions.

Usenet had cancelbots and curated feeds.

Even at the raw IP level, whole networks used to be de-peered if they continued to host illegal (or just awful or inconvenient like spam) content, collateral-damage to co-hosted domains be damned.

There is always some "editing".

Facebook rejects Australia's pay-for-news plan, proposes its own idea: How about no more articles at all, sunshine?

_andrew

Re: Opening up for competition from others

Well I'm sure that Bing (Microsoft) are rubbing their hands in anticipation. The rule only applies to the named companies (Facebook and Google), not to search engines or social media in general. The treasurer has the power to add other companies to the list, but for now that's it.

Of course it won't actually make people use Bing: the default search engine on the built-in browser on the most popular operating system on PCs and never the less everyone goes out of their way to install Chrome and Google search. That has to be telling them something...

_andrew

Re: Klaatu Barata Nikto!

No, not content. Anything that smells like news, from anywhere, that an Australian might read. links shared between friends counts.

The stupid regulation was drafted in full knowledge of what happened in Spain and Germany, and they tried to be clever and write it in a way that would not just result in Australian news being dropped.

Relying on plain-text email is a 'barrier to entry' for kernel development, says Linux Foundation board member

_andrew

Re: Is setting up a non-shitty email client really that hard?

Apple Mail is perfectly capable of sending and displaying plain text email, and I'm fairly sure that the default is the correct option: reply according to the source message type.

The display or non-display of full email addresses in the composition bar is also an option: you could choose to show the whole thing to avoid your (own) confusion. What you see of email addresses when composing or reading has no relationship with what your recipients see: that will be up to their email clients and the settings that they've configured.

Apple Mail does have a couple of infelicities that I'd prefer it didn't: the "load remote content" switch is global; you can't train it that certain senders are OK but most aren't, and so the only safe way to operate (no automatic download of images) requires an extra click for the messages that you do want to see images in. It also has this "guess which outgoing account is most appropriate" mechanism, and I have also had it choose to send work email from my home account, and that in turn has been because I have the default real-name display turned on. It has improved a bit lately, in that it does display your full email address in the From: field when composing, but you still have to be paying attention if it guesses incorrectly.

That said, Apple Mail is still my favorite email client, by far. I never found its equal when I ran a Unix desktop, and don't know of anything comparable on Windows, either.

Australia to force Google and Facebook to pay for news and reveal algorithm changes before they whack web traffic

_andrew

Re: If you require them to tell you exactly what algorithm they use

The corollary of that argument is that similar regulation would be required for all web sites, to ensure that their "internet shop-front signage" was not lying. So trade-practices laws should apply to "SEO", and with any luck put the practice (mostly) out of business.

Seems much more likely that what would really happen is that the game would be over and we'd be back to search results as useful as those of AltaVista.

Be careful what you wish for...

Apple to keep Intel at Arm's length: macOS shifts from x86 to homegrown common CPU arch, will run iOS apps

_andrew

Re: Compatibility is gonna be a problem.

Might be a (small) problem for developers, but it won't be much of a problem for users. Apple has already trained developers that they need to keep up. Their cunning plan, over many years, is to stop un-maintained code from working at all. Ergo: all code that runs at all on a modern mac is actively maintained. And the next version will have Arm-supporting fat binaries.

_andrew

Re: Really?

Recent benchmarks on some sites have put the A13 in the iPhone 11 as faster than the latest MacBook Pro 13-inch. Definitely in the ballpark.

The Register has commented favourably about several server-grade Arm parts being installed by the rack-load in AWS, Azure and GWS, so the "Mac Pro" grade parts should be fine too (Apple's cores are stronger than the ones in those many-core parts).

And for aspiration, Fujitsu just pipped the Supercomputer list by a cool factor of two and a half, entirely based on Arm cores with the new SVE vector instruction set, not a GPU in sight.

Freed from the office, home workers roam sunlit uplands of IPv6... 2 metres apart

_andrew

Re: IPv6 by default

You know that actual firewalls work as a firewall too, it's not just NAT. My shiny new IPv6-capable router from the NBN update had the obvious firewall installed (no incoming connections), just like the NAT default. I'd be very surprised if any ISP's specified or supplied router came with no firewall. That would be mad.

HMD Global pokes head out of quarantine to show off 3 new Nokia mobiles

_andrew

Re: Shipping a new device with MicroUSB?

Never mind the power connector: look at the 28-day standby rating! That's back to what phones used to achieve, before they became "smart" and needed to be charged every night. If true, that's a major achievement, and almost worth giving up 5G and a 64MPixel camera for ;-)

WebAssembly gets nod from W3C and, most likely, an embrace from cryptojackers online

_andrew
Boffin

Re: More secure than Java how?

WASM is very like Java in many respects, but has a significant difference. That difference is very likely why it has taken it as long as it has to gain any kind of traction. The difference is that Java defined a fairly standard sort of standard library, with access to essentially all of the host operating system's resources. Java web applets nominally had a restricted set of APIs available, but they aren't all that restricted, and the big problem is that the libraries are huge, and implemented on top of (not-memory-safe) C and C++ code that turned out to be full of bugs that could be exploited. WASM has no standard libraries, and no object model. Indeed, it has no inherent access to _any_ APIs at all. It's just a blob of code that has access to a pre-allocated chunk of memory. All it can do is run its program when asked by the JavaScript attached to its host web page, which will have to extract the result from a chunk of raw shared memory set up for that purpose beforehand. So the only system access that WASM code has is through the host JavaScript.

At first, host JavaScript just did the specific things that WASM wanted, and it was fairly gnarly. Now though, various groups have gone to the bother of writing fairly sophisticated libraries of interface routines, function-calling mechanisms with argument marshalling and so on, so that fairly normal sorts of developments can in fact be done. But the restrictions are still those of the browser JavaScript, which although not unblemished is not bad at security. Essentially, the sandboxing is significantly stronger this time around.

Page: