But we need a way to identify people, and if you think that driver's licenses and social security numbers aren't already doing this, you're just closing your eyes to it.
There are still some significant differences, though. In particular, I cannot be legally required to produce a driving licence or my National Insurance number in the UK, other than for purposes to which they are relevant (e.g., driving my car, or arranging my financial details). We're still a long way from wanting (or, quite possibly, needing) a universal ID card yet, IMHO.
That said, I totally agree that the privacy vs. identification issue is not a black and white one as too many people make out. That Ben Franklin quote that a certain type of weenie keeps posting to be clever is getting old real fast, and I rather doubt that those posting it would stand up to be first in the firing line when it came down to it.
By the way, I actually did have my identity messed up back at the start of last year, when a government weenie in a tax office mistyped one of those magic numbers. I suddenly worked two full time jobs on opposite sides of the country, and lived in two different places simultaneously. My tax allowances were revoked, without notice, and the first I heard of it was when my pay cheque was short. It took three months to clear up the mess; that started with the tax office refusing to talk to me on the phone because the details they asked for to confirm my identity (my address and current employer) "didn't match what was on their system".
Thanks, but if the government can screw up this spectacularly with the systems it's got in place just now, I'm real happy not to have a legally rqeuired universal ID card, where a similar screw up could interfere with everything I do.
Why don't we just hand over all our biometric data to a trusted third party like microsoft. They could manage the identities of the entire population of the world and free up needed resources for governments.
Perhaps because Microsoft have just last week been given a pretty severe slapping in Europe over the whole Passport/privacy issue, which means they're going to have to behave a whole lot better and give us the opt-outs we want now, maybe?:-)
There was recently a story in the Register (and BBC news) on how there was a large amount of negative feedback using a web-based fax gateway (FaxYourMP.com I think). The government are doing a separate study on this as well, which the stand.org.uk campagn is against. They have received assurances from the government that any web based complaints will be treated as seriously as regular letters of complaint (much easier too).
That's easy to test. Last time I checked, your MP was legally required to respond to every written letter they received from one of their constituents. With due allowance to the fact that one idiot could send a thousand pointless electronic communications in the same minute, will there be a similar legal requirement on MPs to address issues their constituents raise with them via electronic means?
Remember, it was us geeks who convinced the suits that the Internet was the way to travel in the 21st century.
We did? Then perhaps we ought to wake up and admit we were wrong.
A friend was showing me his super new mobile phone earlier this evening. Does picture messaging, has lots of SMS memory, lets you program in polyphonic ring tones, and probably dances on a tabletop if you press * 7 often enough.
But it still took nearly 30 seconds to boot up into a Windows variant. <sigh>
No it's not. Being hit by another vehicle is a life-threatening situation. Downloading pr0n more slowly than usual -- the biggest impact it had on most people, I suspect -- is not.
Investigations from the NTSB and all will force Hyundai to recall all their affected cars and fix the brake problem.
Sure, because there have been absolutely no leaked documents from board level in an Anonymous Major Car Manufacturer in the past, in which the discussion focussed on whether it would be cheaper to recall known dangerous vehicles en masse, or to accept the damages awarded after accidents and subsequent law suits but not fix the cars. No, wait. There have been leaked documents just like that, haven't there?
Is this a joke? I have NEVER had to hack the registry to get an nVidia driver to work at a decent refresh rate.
No, it's not a joke. There are two well-known problems with the 2000/XP series of Windows versions relating to refresh rates. In one case, the refresh rate gets set to the lowest available setting (usually 60Hz), typically due to changes in video mode when playing games. In another, Plug and Play monitors randomly change back into Default monitors, taking away the option of using higher rates with them. These issues are Windows-based not nVidia-specific, though the latter released a tool that fixes some of the problems for users of their cards.
I'm currently writing this using Mozilla on an up-to-date version of Windows XP.
XP fixed most of the edge-case issues here. For example, the Start button... is infinitely wide and deep on the two edges of the screen it shares (bottom-left, top-left, top-right).
No, it's not. My task bar is two rows high, one for app buttons, one for toolbars. The Start button is on the top row, not the bottom one.
IE should be fixed to not allow websites to "maximize" (change the window size to the size of the desktop, rather than actually maximizing) the window, and the super-parent's gripe would go away.
It has nothing to do with IE. Many applications open windows that way. Part of the problem is that Windows allows this pseudo-maximisation without detecting that the window occupies the full desktop (or maybe almost all of it, with some tolerance) and snapping it into true maximised mode.
None of this changes the fact that the rounded corners suck from a usability perspective anyway. Some of us tile windows side-by-side when using two versions of the same program on different documents. Your window isn't maximised, but you still have cornering issues.
But white looks cleaner and more modern, so website designers normally use it.
A lot has to do with accepted norms, I think. I've seen a couple of surveys (sorry, no links off the top of my head) that looked at the top 10 or 20 most visited web sites and tried to identify common features. A couple of biggies were the fact that almost all of them used small to medium text in black on white, often in some sort of sidebar layout with the main content in the middle and links, search boxes etc. on the left and/or right.
Obviously, it doesn't prove a causal relationship, but if you can rely on almost anyone familiar with the web to have visited at least a couple of these sites fairly regularly (and you probably can) then they do a lot towards defining the accepted norm, and thus what people perceive to be easy to use, good style, etc.
My biggest problem with the Windows XP theme is the rounded window "corners". Often times, an IE window looks maximized, but because of the few missing pixels on the rounded edge, I end up hitting the X/Close button of the IE window *below* the top-most window.
Yep, that one's a serious screw-up. Ask any usability guy about where you can fastest reach with the mouse, and he'll tell you that it's exactly where it is, then the four corners of the screen, then the edges of the screen. It's much slower to move to somewhere slightly off an edge, or slightly away from the current position, because you don't have the inherent error correction in where the pointer winds up. Thus, having controls (including menus, buttons or whatever) just off the edges of the screen but not reaching right to the edge is a Usability Sin(TM) and should be banned. I'm amazed Microsoft's usability guys missed this one, and that they haven't yet fixed it...
Copyright violation right now has the same stigma of speeding. You know its dangerous. You know it's against the law. You also know that *Everyone* does it. Even lawmakers.
One of the sentences above is demonstrably incorrect most of the time, in both cases. Any takers?
I remember reading that John C is going to cap the FPS on Doom3 to like 30 or 40 FPS per seconds.
That seems unlikely, since first person shooters like Doom III are exactly where humans do often perceive a difference between 30+ and 50+ or 70+ average FPS. Granted if he's capping the actual FPS figure and not the average that's not entirely the same thing, but nonetheless, it seems a very odd thing to do. Do you have any source to support this?
Unfortunately, I had no success at first with any combination of Cat 2.4, 2.5 or 3.0, or the Crucial-supplied drivers for the card, with DX9. I tried some of them with DX 8.1 (as shipped with WinXP) as well, with similar problems.
I think I may finally have got it sorted just recently, having reinstalled Catalyst 3.0 drivers after DirectX 9. (There's a tip somewhere in the ATI info that suggests this.) At least I've got both installed and things like DirectX diagnostics and the DX9 demo screensavers don't appear to crash out any more. I still have some stability problems (very new PC) so I'm reserving judgement but crossing my fingers at this point...
I just bought a brand new Radeon 9700 Pro for my brand new PC. Eventually it will dual boot WinXP/Linux, but I'm waiting for the next version of a particular distro, so for the immediate future it's XP only.
You know what? ATI's Windows drivers still appear to suck. Contrary to what I'd read before buying, their new Catalyst 3 drivers do not seem to be completely stable with DirectX 9 (their raison d'etre). I got copies of both Neverwinter Nights and Black and White for Christmas, both games several of my friends have enjoyed playing and things I was looking forward to. Right now I (and a lot of other people with ATI cards, apparently) can't play NWN for more than five minutes without it taking out my whole PC.
If Linux has ATI drivers that work at all with the advanced features on the graphics cards, it's going to be better than Windows.
(I put together the PC myself, BTW, from nothing but well-regarded and highly recommended kit. The only other drivers I've installed apart from the ATI ones and DirectX 9 are those supplied with the mobo to handle the on-board kit. I'm thus reasonably confident that it really is the combination of Catalyst drivers and DirectX that accounts for the lack of stability.)
You should get a clause signed that any work you do on GPL or Open Source for yourself or the company can not be claimed by the company as well.
Stuff that. You should be looking for a clause that says your employer has no rights whatsoever to anything you ever do that doesn't involve at least one of company time and company resources.
GPL, open source or whatever is totally irrelevant.
Everybody says that with their mouth, but not with their wallets.
I disagree. I just bought a good ol' Seagate Barracuda IV, first and foremost because amongst other experienced home PC builders in the area, there have been lots of problems with every other make in the past year or two, but never a complaint about this drive AFAIK. It's slower than its rivals, probably cost slightly more, and isn't as big. It's also fast enough for me, big enough for me, and hopefully reliable enough to last as long as the PC it's going in. The warranty still ain't all that, which is annoying, but I've definitely put expected reliability before either price or performance.
Have you ever tried [pair programming] seriously? If not, you are falling into the trap of speaking in ignorance.
Don't knock it 'til you try it.
With all due respect, why do XP evangelists always assume that those of us who disagree with their golden rules either have never tried them or have failed to understand or appreciate them?
There are plenty of good, experienced programmers who have experimented with various aspects of XP, including the big ones of pair programming and a test-first process, and have found them to be inappropriate or ineffective for them. If they work for you, that's great and I'll be the last person to tell you you should do anything else, but that's you, not me.
Personally, I can see the merit in these approaches, but I don't hold them on a pedestal as some sort of panacea. I reckon my programming matches the 80%/20% pattern pretty well. In my particular circumstances, when I get hung up on something, it's usually because I'm not aware of a relevant detail in how our software works, being a relatively new member of the team. Speaking to someone more experienced to find the missing link, or bouncing an idea off another team member to see where they instinctively see it going, is often a way to get clear of the 80% and back into the 20% again for me, right now. Given how naturally our group works together anyway, I could see pair programming potentially working out.
On the other hand, we have a very effective lightweight process going on, which is based on an initial proposal for a feature, followed by a very iterative design and implementation phase, with automated tests. I wouldn't swap that for a test-first process given any choice in the matter, because I've tried both, and found the latter lacks the power to work well in my development areas. If it works for you, that's great, please use it, but don't assume that it would work for everyone else just as well...
You are NOT overriding the rules of the language. A non const value can ALWAYS legally be made const.
You're not overriding it in that sense, sure. My point was that the language rules are clear on the overriding: if you call the method on a non-const object, you get the non-const method, whether or not you will use it as such. This can be a pain, for example in the context of operator[] on an associative array type, where you might prefer to throw an exception rather than implicitly create an entry that does not yet exist if you're looking up rather than assigning. You are, of course, at liberty to convert the returned value yourself into a constant form, but that is a second step that the language would not ordinarily take for you.
The purpose of something like auto would be to replace the first step. If you're wanting to explicitly set the type, you don't need auto. Its value is elsewhere; look at some of the recent template discussions in places like comp.lang.c++.moderated and you'll find the need for that feature, or something like typeof, does come up every now and then.
I appreciate that it's not always what you want, but it is always what you'll get. When you use a const_iterator on a non-const container, you're still getting a normal iterator back first, it's just then being implicitly converted to the constant access version. If you want to do that, you're explicitly overriding the rules of the language, in which case presumably the auto stuff mentioned isn't what you're looking for anyway.
OK, I give up. You claim to be concerned about performance, you worry about cache misses and indirections, yet you ignore one of the most common but subtle performance issues of most contemporary architectures. You obviously don't understand the issue of padding on modern architectures: we're talking about 32-bit boundaries on 32-bit systems here, not byte alignment, and the reason we're talking about that is the slow access times to memory that is not so aligned. The QoI issue with compilers is with those that don't align this way for good performance where appropriate, not with those that do.
You maintain that using indirection can cause a "severalfold space and time overhead" when on most systems, it only doubles the space overhead even in the worst possible case. Looking it up involves a single extra indirection, and that only the first time under typical conditions, which is hardly a "severalfold" performance hit.
And you tell me all about how no-one has solved this problem well enough in thirty years, yet Java is sitting here as one of the most successful (in terms of industry penetration, volume of code written, etc.) languages of the past decade.
You tell me that you write code that's portable, but tuned only for GNU C++. If performance mattered to you that much, you'd surely be aware that the price you pay for keeping GCC portable is that its output code can run several times slower than that generated by a compiler that's aware of the issues we've discussed and accounts for them properly.
Finally, you belittle the job that I do and attack me personally over my knowledge of this subject, in spite of the fact that these issues are directly relevant to me, day in and day out, whereas you clearly aren't working in a performance-driven environment at all.
If it pleases you to post such comments, you go right ahead, but it doesn't make your facts any more correct or your arguments any more sensible, no matter how many times you repeat yourself, how often you provide demonstrably incorrect information, or how often you insult the person disagreeing with you.
The facts remain that while value types are certainly useful, and faster than reference types, they are not essential for most work, and not having them is not a sufficient reason to rule out a particular programming language for most developments.
The context is (and must be) unambiguous at the time of the function call. In this case, you'll get the const version if you call begin() on a const object, and it's basically as simple as that. Whether or not you'll ever modify data doesn't have any bearing on its constancy in C++, which is unfortunate at other times, but helpful here.
You can resolve the ambiguity using the same rules that decide which version you'll be calling (which you have to work out anyway, of course) and take the return type of that version. As long as you've got context -- which you have in the example case -- it's not a problem. In general, you'd need to specify which overload you wanted. It's kind of like explicit template instantiation (and equally horrible if you have to do it).
Slashdot Mirror
There are still some significant differences, though. In particular, I cannot be legally required to produce a driving licence or my National Insurance number in the UK, other than for purposes to which they are relevant (e.g., driving my car, or arranging my financial details). We're still a long way from wanting (or, quite possibly, needing) a universal ID card yet, IMHO.
That said, I totally agree that the privacy vs. identification issue is not a black and white one as too many people make out. That Ben Franklin quote that a certain type of weenie keeps posting to be clever is getting old real fast, and I rather doubt that those posting it would stand up to be first in the firing line when it came down to it.
By the way, I actually did have my identity messed up back at the start of last year, when a government weenie in a tax office mistyped one of those magic numbers. I suddenly worked two full time jobs on opposite sides of the country, and lived in two different places simultaneously. My tax allowances were revoked, without notice, and the first I heard of it was when my pay cheque was short. It took three months to clear up the mess; that started with the tax office refusing to talk to me on the phone because the details they asked for to confirm my identity (my address and current employer) "didn't match what was on their system".
Thanks, but if the government can screw up this spectacularly with the systems it's got in place just now, I'm real happy not to have a legally rqeuired universal ID card, where a similar screw up could interfere with everything I do.
Perhaps because Microsoft have just last week been given a pretty severe slapping in Europe over the whole Passport/privacy issue, which means they're going to have to behave a whole lot better and give us the opt-outs we want now, maybe? :-)
That's easy to test. Last time I checked, your MP was legally required to respond to every written letter they received from one of their constituents. With due allowance to the fact that one idiot could send a thousand pointless electronic communications in the same minute, will there be a similar legal requirement on MPs to address issues their constituents raise with them via electronic means?
We did? Then perhaps we ought to wake up and admit we were wrong.
A friend was showing me his super new mobile phone earlier this evening. Does picture messaging, has lots of SMS memory, lets you program in polyphonic ring tones, and probably dances on a tabletop if you press * 7 often enough.
But it still took nearly 30 seconds to boot up into a Windows variant. <sigh>
No it's not. Being hit by another vehicle is a life-threatening situation. Downloading pr0n more slowly than usual -- the biggest impact it had on most people, I suspect -- is not.
Sure, because there have been absolutely no leaked documents from board level in an Anonymous Major Car Manufacturer in the past, in which the discussion focussed on whether it would be cheaper to recall known dangerous vehicles en masse, or to accept the damages awarded after accidents and subsequent law suits but not fix the cars. No, wait. There have been leaked documents just like that, haven't there?
Nah, that was a Freudian slap. A Freudian slip is when you say one thing and mean your mother.
No, it's not a joke. There are two well-known problems with the 2000/XP series of Windows versions relating to refresh rates. In one case, the refresh rate gets set to the lowest available setting (usually 60Hz), typically due to changes in video mode when playing games. In another, Plug and Play monitors randomly change back into Default monitors, taking away the option of using higher rates with them. These issues are Windows-based not nVidia-specific, though the latter released a tool that fixes some of the problems for users of their cards.
I'm currently writing this using Mozilla on an up-to-date version of Windows XP.
No, it's not. My task bar is two rows high, one for app buttons, one for toolbars. The Start button is on the top row, not the bottom one.
It has nothing to do with IE. Many applications open windows that way. Part of the problem is that Windows allows this pseudo-maximisation without detecting that the window occupies the full desktop (or maybe almost all of it, with some tolerance) and snapping it into true maximised mode.
None of this changes the fact that the rounded corners suck from a usability perspective anyway. Some of us tile windows side-by-side when using two versions of the same program on different documents. Your window isn't maximised, but you still have cornering issues.
A lot has to do with accepted norms, I think. I've seen a couple of surveys (sorry, no links off the top of my head) that looked at the top 10 or 20 most visited web sites and tried to identify common features. A couple of biggies were the fact that almost all of them used small to medium text in black on white, often in some sort of sidebar layout with the main content in the middle and links, search boxes etc. on the left and/or right.
Obviously, it doesn't prove a causal relationship, but if you can rely on almost anyone familiar with the web to have visited at least a couple of these sites fairly regularly (and you probably can) then they do a lot towards defining the accepted norm, and thus what people perceive to be easy to use, good style, etc.
Yep, that one's a serious screw-up. Ask any usability guy about where you can fastest reach with the mouse, and he'll tell you that it's exactly where it is, then the four corners of the screen, then the edges of the screen. It's much slower to move to somewhere slightly off an edge, or slightly away from the current position, because you don't have the inherent error correction in where the pointer winds up. Thus, having controls (including menus, buttons or whatever) just off the edges of the screen but not reaching right to the edge is a Usability Sin(TM) and should be banned. I'm amazed Microsoft's usability guys missed this one, and that they haven't yet fixed it...
One of the sentences above is demonstrably incorrect most of the time, in both cases. Any takers?
That seems unlikely, since first person shooters like Doom III are exactly where humans do often perceive a difference between 30+ and 50+ or 70+ average FPS. Granted if he's capping the actual FPS figure and not the average that's not entirely the same thing, but nonetheless, it seems a very odd thing to do. Do you have any source to support this?
Unfortunately, I had no success at first with any combination of Cat 2.4, 2.5 or 3.0, or the Crucial-supplied drivers for the card, with DX9. I tried some of them with DX 8.1 (as shipped with WinXP) as well, with similar problems.
I think I may finally have got it sorted just recently, having reinstalled Catalyst 3.0 drivers after DirectX 9. (There's a tip somewhere in the ATI info that suggests this.) At least I've got both installed and things like DirectX diagnostics and the DX9 demo screensavers don't appear to crash out any more. I still have some stability problems (very new PC) so I'm reserving judgement but crossing my fingers at this point...
I just bought a brand new Radeon 9700 Pro for my brand new PC. Eventually it will dual boot WinXP/Linux, but I'm waiting for the next version of a particular distro, so for the immediate future it's XP only.
You know what? ATI's Windows drivers still appear to suck. Contrary to what I'd read before buying, their new Catalyst 3 drivers do not seem to be completely stable with DirectX 9 (their raison d'etre). I got copies of both Neverwinter Nights and Black and White for Christmas, both games several of my friends have enjoyed playing and things I was looking forward to. Right now I (and a lot of other people with ATI cards, apparently) can't play NWN for more than five minutes without it taking out my whole PC.
If Linux has ATI drivers that work at all with the advanced features on the graphics cards, it's going to be better than Windows.
(I put together the PC myself, BTW, from nothing but well-regarded and highly recommended kit. The only other drivers I've installed apart from the ATI ones and DirectX 9 are those supplied with the mobo to handle the on-board kit. I'm thus reasonably confident that it really is the combination of Catalyst drivers and DirectX that accounts for the lack of stability.)
Stuff that. You should be looking for a clause that says your employer has no rights whatsoever to anything you ever do that doesn't involve at least one of company time and company resources.
GPL, open source or whatever is totally irrelevant.
Well, someone around here has certainly been popping a pill, but I'm not sure it was the original poster... :-)
I disagree. I just bought a good ol' Seagate Barracuda IV, first and foremost because amongst other experienced home PC builders in the area, there have been lots of problems with every other make in the past year or two, but never a complaint about this drive AFAIK. It's slower than its rivals, probably cost slightly more, and isn't as big. It's also fast enough for me, big enough for me, and hopefully reliable enough to last as long as the PC it's going in. The warranty still ain't all that, which is annoying, but I've definitely put expected reliability before either price or performance.
With all due respect, why do XP evangelists always assume that those of us who disagree with their golden rules either have never tried them or have failed to understand or appreciate them?
There are plenty of good, experienced programmers who have experimented with various aspects of XP, including the big ones of pair programming and a test-first process, and have found them to be inappropriate or ineffective for them. If they work for you, that's great and I'll be the last person to tell you you should do anything else, but that's you, not me.
Personally, I can see the merit in these approaches, but I don't hold them on a pedestal as some sort of panacea. I reckon my programming matches the 80%/20% pattern pretty well. In my particular circumstances, when I get hung up on something, it's usually because I'm not aware of a relevant detail in how our software works, being a relatively new member of the team. Speaking to someone more experienced to find the missing link, or bouncing an idea off another team member to see where they instinctively see it going, is often a way to get clear of the 80% and back into the 20% again for me, right now. Given how naturally our group works together anyway, I could see pair programming potentially working out.
On the other hand, we have a very effective lightweight process going on, which is based on an initial proposal for a feature, followed by a very iterative design and implementation phase, with automated tests. I wouldn't swap that for a test-first process given any choice in the matter, because I've tried both, and found the latter lacks the power to work well in my development areas. If it works for you, that's great, please use it, but don't assume that it would work for everyone else just as well...
You're not overriding it in that sense, sure. My point was that the language rules are clear on the overriding: if you call the method on a non-const object, you get the non-const method, whether or not you will use it as such. This can be a pain, for example in the context of operator[] on an associative array type, where you might prefer to throw an exception rather than implicitly create an entry that does not yet exist if you're looking up rather than assigning. You are, of course, at liberty to convert the returned value yourself into a constant form, but that is a second step that the language would not ordinarily take for you.
The purpose of something like auto would be to replace the first step. If you're wanting to explicitly set the type, you don't need auto. Its value is elsewhere; look at some of the recent template discussions in places like comp.lang.c++.moderated and you'll find the need for that feature, or something like typeof, does come up every now and then.
We know, thanks.
Love and hugs,
The RIAA
Close.
s/Kill/Sue/
I appreciate that it's not always what you want, but it is always what you'll get. When you use a const_iterator on a non-const container, you're still getting a normal iterator back first, it's just then being implicitly converted to the constant access version. If you want to do that, you're explicitly overriding the rules of the language, in which case presumably the auto stuff mentioned isn't what you're looking for anyway.
OK, I give up. You claim to be concerned about performance, you worry about cache misses and indirections, yet you ignore one of the most common but subtle performance issues of most contemporary architectures. You obviously don't understand the issue of padding on modern architectures: we're talking about 32-bit boundaries on 32-bit systems here, not byte alignment, and the reason we're talking about that is the slow access times to memory that is not so aligned. The QoI issue with compilers is with those that don't align this way for good performance where appropriate, not with those that do.
You maintain that using indirection can cause a "severalfold space and time overhead" when on most systems, it only doubles the space overhead even in the worst possible case. Looking it up involves a single extra indirection, and that only the first time under typical conditions, which is hardly a "severalfold" performance hit.
And you tell me all about how no-one has solved this problem well enough in thirty years, yet Java is sitting here as one of the most successful (in terms of industry penetration, volume of code written, etc.) languages of the past decade.
You tell me that you write code that's portable, but tuned only for GNU C++. If performance mattered to you that much, you'd surely be aware that the price you pay for keeping GCC portable is that its output code can run several times slower than that generated by a compiler that's aware of the issues we've discussed and accounts for them properly.
Finally, you belittle the job that I do and attack me personally over my knowledge of this subject, in spite of the fact that these issues are directly relevant to me, day in and day out, whereas you clearly aren't working in a performance-driven environment at all.
If it pleases you to post such comments, you go right ahead, but it doesn't make your facts any more correct or your arguments any more sensible, no matter how many times you repeat yourself, how often you provide demonstrably incorrect information, or how often you insult the person disagreeing with you.
The facts remain that while value types are certainly useful, and faster than reference types, they are not essential for most work, and not having them is not a sufficient reason to rule out a particular programming language for most developments.
The context is (and must be) unambiguous at the time of the function call. In this case, you'll get the const version if you call begin() on a const object, and it's basically as simple as that. Whether or not you'll ever modify data doesn't have any bearing on its constancy in C++, which is unfortunate at other times, but helpful here.
You can resolve the ambiguity using the same rules that decide which version you'll be calling (which you have to work out anyway, of course) and take the return type of that version. As long as you've got context -- which you have in the example case -- it's not a problem. In general, you'd need to specify which overload you wanted. It's kind of like explicit template instantiation (and equally horrible if you have to do it).