Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


IPv6 - What is it good for? - Page 2
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

IPv6 - What is it good for?

2

Comments

  • @WebProject said:

    @user54321 said:

    @WebProject said:
    Some ISP providers do support some still don't know anything about it, example:
    EE mobile - IPv6 enabled
    EE broadband - No
    Vodafone Mobile - No
    Three Network - No
    BT - partially using

    like i said idiots that can't be bothered, they all could support IPv6 if they want to, the equipment they use does support it.

    they do have valid excuses - cost and not all devices do support it!

    Like today it rained so we won't deploy IPv6 for the next 20 years or what? So you want that ICANN sets a deadline and they will remove all IPv4 allocations at 6/6/2020?

  • @ITLabs said:
    For making fancy mugs.

    or or or

    "> @TimboJones said:

    @user54321 said:

    @WebProject said:
    Some ISP providers do support some still don't know anything about it, example:
    EE mobile - IPv6 enabled
    EE broadband - No
    Vodafone Mobile - No
    Three Network - No
    BT - partially using

    like i said idiots that can't be bothered, they all could support IPv6 if they want to, the equipment they use does support it.

    Sounds like you're not familiar with how most businesses are run. If they don't NEED to spend the money, effort and resources, don't. Being able to put off a decision as long as possible is generally the best approach.

    I mean, for the Canadian ISPs with IPv4 subnets coming out their asses, they avoided at least three failed early IPv6 implementations and all the support hassle it would have caused.

    coughs bell

    To be fair, they added native V6 to their mobile/data network a few months ago. Still waiting for V6 on fibre...

    Note: Rogers got native V6 on broadband a while back iirc

    Thanked by 1ITLabs
  • jsgjsg Member, Resident Benchmarker
    edited July 2019

    @user54321 said:
    Sure go ahead with IPv6, everything works, the only problem are idiots that don't support it. If everybody would just implemented it no matter if they needed or not IPv4 would be already dead for 20 years, but there is always somebody who thinks, I don't need it so i don't support it. I do it now the same way and removed IPv4 support for many of my services because and give a fuck about users without IPv6.

    You obviously do not understand how both business and networking works.
    No sane company just implements something, no matter wther they need it or not, especially if implementing it does not mean to flick a switch or to spend some hundred bucks but lots of money and lots of problems and risks.

    Plus anyone with a working brain asks first "Why and what for?". The common answer usually was that we are running out of IP4 addresses. Soon. Frighteningly soon.

    Well that turned out to be wrong for many, many years - which is no surprise considering where ISPs are -> right between millions of customers and the internet. So of course they first asked how they could keep running their operation with a cheaper and less grave solution.

    Just in case you really want to understand what you are talking about: the problem never really was that IP4 doesn't provide enough addresses. The real problem was a lack of thinking and an attitude that created lots and lots of wasting. Wasting as in throwing a /16 at schools or as in server software that could serve only one single domain. DNS not providing port information is another good example for a lack of thinking that lead to waste.

  • user54321user54321 Member
    edited July 2019

    @jsg i don't care about what ppl think, the normal process is you deprecate stuff after it got replaced by something, that worked for pretty much anything that got a internet standard, only IPv4 simply won't die. And it should have 2000 died.

  • @user54321
    Surely it's a bit of a platitude to expect an entire world's worth of hardware and software to seamlessly switch to another protocol... for an ideal. Although IPv4 is scarce I've never heard of that fact preventing people from doing anything meaningful - probably more the opposite in fact, prevents people from doing non-meaningful stuff without paying a few bucks for an IP.

    In the end someone has to pay, whether it's for the jacked rates of IPv4, implementing IPv6, or paying someone who paid for it.

    Thanked by 1uptime
  • hzrhzr Member

    ricardo said: Surely it's a bit of a platitude to expect an entire world's worth of hardware and software to seamlessly switch to another protocol... for an ideal.

    It has been multiple decades

    Thanked by 1skorous
  • user54321user54321 Member
    edited July 2019

    guys internet standards don't work like a company, if they would there would be no internet anymore.

  • ricardoricardo Member
    edited July 2019

    hzr said: It has been multiple decades

    With no inherent need to upgrade. See the previous post.

    It's either market forces or law that changes things, not ideals about having 2^xxx IPs for every atom in the universe.

    In any event I prefer the previously mentioned idea of a 64-bit address, much more friendly for processing and more than enough.

    user54321 said: guys internet standards don't work like a company, if they would there would be no internet anymore.

    Bit of a vacuous statement, shirley.

  • @hzr said:

    ricardo said: Surely it's a bit of a platitude to expect an entire world's worth of hardware and software to seamlessly switch to another protocol... for an ideal.

    It has been multiple decades

    And it would be another two decades if not for the billions the US government put into vendors pockets to subsidize and pay for R&D and development. You can ask many of the network players how much and how many customers cared about IPv6 last decade. I remember Ixia guy telling me the US government throwing money at them for years and pretty much only them.

  • jsgjsg Member, Resident Benchmarker

    @user54321 said:
    @jsg i don't care about what ppl think,

    How then do you understand their decisions?

    the normal process is you deprecate stuff after it got replaced by something, that worked for pretty much anything that got a internet standard, only IPv4 simply won't die. And it should have 2000 died.

    IP4 -> IPv6 is not like a car for a taxi business that gets replaced. Changing to IPv6 would be much much more and much more complicated than simply changing equipment.

    Btw. IPv6 itself has seen quite some changes in those 20 years - and for a reason I guess.

    @user54321 said:
    guys internet standards don't work like a company, if they would there would be no internet anymore.

    That's simply BS. If you ever looked how internet standards come to life you would see that most of them are with very heavy involvement of (usually large) corporations - in part for sensible reasons and in part for not so sensible ones.

    Thanked by 1uptime
  • @jsg said:

    @user54321 said:
    @jsg i don't care about what ppl think,

    How then do you understand their decisions?

    most of them are based on greed so it is a waste of time.

    the normal process is you deprecate stuff after it got replaced by something, that worked for pretty much anything that got a internet standard, only IPv4 simply won't die. And it should have 2000 died.

    IP4 -> IPv6 is not like a car for a taxi business that gets replaced. Changing to IPv6 would be much much more and much more complicated than simply changing equipment.

    Btw. IPv6 itself has seen quite some changes in those 20 years - and for a reason I guess.

    so you still telnet in your servers?

    @user54321 said:
    guys internet standards don't work like a company, if they would there would be no internet anymore.

    That's simply BS. If you ever looked how internet standards come to life you would see that most of them are with very heavy involvement of (usually large) corporations - in part for sensible reasons and in part for not so sensible ones.

    So the goal of a internet standard is infinite growth? If the consensus is reached wich it has with publishing the rfc there is no goal beyond that. There is only greed of companys to save some cents and shit on their customers with not enabling it. Nobody needs to write implementations for IPv6 anymore they just need to adopt their configs.

    Thanked by 1skorous
  • jsgjsg Member, Resident Benchmarker

    @user54321 said:
    most of them are based on greed so it is a waste of time.

    I don't think that most decisions of the people are based on greed. Stupidity comes to mind as a far more important factor but there are many more.

    so you still telnet in your servers?

    I didn't say that I am against real progress - which IPv6 is not.

    So the goal of a internet standard is infinite growth?

    No. Growth is certainly one factor, btw. in academia often too, but there are more factors. And obviously not all business segments are in agreement; it just so happened that the pro IPv6 won this time, maybe even because committees (incl. the business interests there) can decide a lot of funny things but still they decide only how something is to be implemented if it is implemented - that however, if it's done, isn't decided in standards committees (but in board rooms).

    Btw, you are implying (and alleging) a lot here, e.g. "business == desire for infinite growth (or greed)" There are counter examples so your logic is flawed.

  • Wow, perhaps 10 years later, people will still be arguing with each others why IPv6 is needed. :)

  • jsgjsg Member, Resident Benchmarker
    edited August 2019

    @try4lontalk said:
    Wow, perhaps 10 years later, people will still be arguing with each others why IPv6 is needed. :)

    No. What I am discussing is

    • do we really need a successor to IP4?
    • and if we do need one then what requirements would that successor need to meet?

    Re the first question: Probably yes, sooner or later, but the real reasons is mostly not (yet) a real need but rather the long time wasting of IP4.
    Re the second question: Funnily (and sadly and shockingly) we still haven't thought a lot about that. Basically all we have so far is what a bunch of idiots came up with plus hysterically "convinced" proponents of their "solution" (IPv6).

    Real engineers would look at and think about both questions and they would keep feasibility in mind (like "is that reasonable with todays processors?). Plus they would have a well founded basis for their numbers rather than offering BS. Finally and importantly real engineers would come up with a solution which sensibly extends what we have now, among other reasons to make changing from IP4 to IPx easy and practical. One typical scheme to do that is to include the current IP4 address space a "region zero" of the new one.
    Example (simple, stupid, just for the sake of explanation): IPx adds 4 bytes to the current 4 bytes (total len 64 bits) but (a) only uses the lower 2 new bytes (the upper 2 bytes being zero) for some time, and (b) '0.0.0.0.a.b.c.d' would be the current IP4 address space and would be perfectly valid. All existing network equipment could work with that and only core (major backbone) equipment would need to be replaced/extended to handle full IPx (all 64 bits).

    Plus smart people would learn from the grave errors that were made with IP4 (the wasting) and would develop reasonable hand out rules like "everybody asking for it can get 16 IPs but to get more you have to provide a sensible case". Additionally it would seem reasonable to have some partitioning of the address space like 0.a.b.c.d to 15.a.b.c.d (16 times the total current address space!) is "wild" (no partitioning) and all higher ranges are partitioned in purpose (e.g. "servers", "end users", "phone") as well as regional spaces (in a fair and reasonable scheme and not like back then like "plenty for the USA and then some more, and the small rest for the whole world"). China and India, for example obviously would get a much bigger enduse space than the USA.

    Well, whatever, one thing is clear: IPv6 is not a reasonable answer to the relevant questions but the result of too much drugs in the mental asylum.

  • jsgjsg Member, Resident Benchmarker

    @perryoo11 said:

    Bad image. It should say "IPv6" and the guy shouldn't say "to keep her running" but "to make her running".

  • @jsg said:

    @perryoo11 said:

    Bad image. It should say "IPv6" and the guy shouldn't say "to keep her running" but "to make her running".

  • jsgjsg Member, Resident Benchmarker

    @user54321

    I disagree with you re IPv6 but I think that you are neither stupid nor trolling.

  • @jsg said:
    @user54321

    I disagree with you re IPv6 but I think that you are neither stupid nor trolling.

    your argument that IPv6 would be more expensive is BS, or is a IPv4 only VPS cheaper than a NAT one with IPv6, no? There are even hosters like scaleway that have IPv6 only VPS which charge you for IPv4 from the first address.
    Have you ever used IPv6?

  • @user54321 said:

    @jsg said:
    @user54321

    I disagree with you re IPv6 but I think that you are neither stupid nor trolling.

    your argument that IPv6 would be more expensive is BS, or is a IPv4 only VPS cheaper than a NAT one with IPv6, no? There are even hosters like scaleway that have IPv6 only VPS which charge you for IPv4 from the first address.
    Have you ever used IPv6?

    Dude, you are so clueless. All the companies who make networking products would have increased costs in software and hardware development. A shit ton of money. QA work more than doubles. I worked for a company that made wimax radios and when the costs to add IPv6 support was mentioned to the customer, they said they didn't need it. So if it's free, they want it. If they need to pay for development, they don't.

  • user54321user54321 Member
    edited August 2019

    so a wimax radio is now a ISP? thanks for the info, that was new to me....

  • jsgjsg Member, Resident Benchmarker

    @user54321 said:
    your argument that IPv6 would be more expensive is BS, or is a IPv4 only VPS cheaper than a NAT one with IPv6, no? There are even hosters like scaleway that have IPv6 only VPS which charge you for IPv4 from the first address.
    Have you ever used IPv6?

    (a) On what is your statement (IPv6 is not more expensive) based?
    (b) Is the type of IP the only or even the decisve factor in VPS pricing?
    (c) What does "a provider hands out IPv6 for free" prove?
    (d) There are (erring at my disadvantage) about 5 Mio packets that go through a 40 Gb/s line (which is by no means a "large" line nowadays) which means that the processor in a router has far less than 1 microsecond to process a packet at wire speed. Based on currently used processors it is a very major difference whether that processing happens within the word size of the processor or not.
    Note: 4 times the word size does not mean 1/4 of the speed but 1/10 if one is very lucky. Add to that the fact that the lookup tables also grow a lot which is a double disadvantage because that memory must be read/written and doing that is typically about a factor 100 slower than on die memory (caches).
    TL;DR IPv6 is dramatically increasing cost and not just in dollars.

    And while that is a major factor it is not the only one. One example for additional cost factors is the need to train personel because the IPv6 people didn't care about ease of migration.

    Thanked by 1TimboJones
  • @jsg said:

    @user54321 said:
    your argument that IPv6 would be more expensive is BS, or is a IPv4 only VPS cheaper than a NAT one with IPv6, no? There are even hosters like scaleway that have IPv6 only VPS which charge you for IPv4 from the first address.
    Have you ever used IPv6?

    (a) On what is your statement (IPv6 is not more expensive) based?

    on price tags.

    (b) Is the type of IP the only or even the decisve factor in VPS pricing?

    on lowend yes, because IPv4 costs money.

    (c) What does "a provider hands out IPv6 for free" prove?

    That there is enough IPv6 space to do this, while IPv4 is very limited.

    (d) There are (erring at my disadvantage) about 5 Mio packets that go through a 40 Gb/s line (which is by no means a "large" line nowadays) which means that the processor in a router has far less than 1 microsecond to process a packet at wire speed. Based on currently used processors it is a very major difference whether that processing happens within the word size of the processor or not.

    you don't do that on a processor, there are offloading features to do that on the NIC, you would max out your processor with only the interrupts caused by doing it in software already at 1,5 GBit/s no matter if IPv4 or IPv6.

    You should ask ISPs what they pay for the machines to do carrier grade nat because they have more customers than IPv4 space. And than tell me again that IPv4 is cheaper lol.

    Note: 4 times the word size does not mean 1/4 of the speed but 1/10 if one is very lucky. Add to that the fact that the lookup tables also grow a lot which is a double disadvantage because that memory must be read/written and doing that is typically about a factor 100 slower than on die memory (caches).
    TL;DR IPv6 is dramatically increasing cost and not just in dollars.

    I'm sorry but it doesn't seem like you have a clue about what you talk there.

    And while that is a major factor it is not the only one. One example for additional cost factors is the need to train personel because the IPv6 people didn't care about ease of migration.

    So it comes again down to stupidity and greed. And with that I have enough time wasted on you.

  • jsgjsg Member, Resident Benchmarker
    edited August 2019

    @user54321 said:
    on price tags.

    Or more precisely: on an arbitrary selection of price tags.

    (b) Is the type of IP the only or even the decisve factor in VPS pricing?

    on lowend yes, because IPv4 costs money.

    Provably wrong.

    That there is enough IPv6 space to do this, while IPv4 is very limited.

    Wrong because the model is quite different. With IP4 the (expensive) equipment, training, etc. is already payed off but the IP ranges are expensive. With IPv6 the IP ranges are cheap but everything else is expensive.

    you don't do that on a processor, there are offloading features to do that on the NIC, you would max out your processor with only the interrupts caused by doing it in software already at 1,5 GBit/s no matter if IPv4 or IPv6.

    Uhm, and on a Nic there is what that does the work? A processor.
    But you are wrong anyway because the offloaded work that can be and is done on the Nic (which btw. is a PC device and not a router or switch device) is not routing.

    You should ask ISPs what they pay for the machines to do carrier grade nat because they have more customers than IPv4 space. And than tell me again that IPv4 is cheaper lol.

    You obviously do not even know how it's done.

    Note: 4 times the word size does not mean 1/4 of the speed but 1/10 if one is very lucky. Add to that the fact that the lookup tables also grow a lot which is a double disadvantage because that memory must be read/written and doing that is typically about a factor 100 slower than on die memory (caches).
    TL;DR IPv6 is dramatically increasing cost and not just in dollars.

    I'm sorry but it doesn't seem like you have a clue about what you talk there.

    Talking to yourself in front of a mirror? Sorry but I provided qualified arguments and your comment is painfully sub-par and actually suggests that you lack the knowledge to offer a real argument. Try that with idiots but not with someone who earns a living by knowing how to optimize throughput.

    So it comes again down to stupidity and greed. And with that I have enough time wasted on you.

    No, your conclusion is wrong (no surprise there). As for your "escape", once more: Don't try tricks that work in your peer group with adults.

    In summary I have seen very little in actual arguments and lots of blabla, belief based defense and propaganda attempts and personal attacks - none of which is able to replace facts and verifiable arguments.

  • jsgjsg Member, Resident Benchmarker

    @user54321 said:
    dude first read [some academic paper from some frankly insignificant university]

    Well, probably the problem is that unlike you I actually understand that paper because I actually worked with tries (I love them but like all devices they have strong as well as weak points).
    For a start: funny tricks with tries are not at all new, neither is that one. Men have invented many ways to massage them and I myself once developped a strange one for a somewhat related use case (a firewall)
    Another problem is that the authors made multiple errors. To name a grave one, they state that 100 Gb/s are 250 M packets worst case. That's wrong and far off. The worst case is all UDP packets for which the number of pps is almost double which means a processing time of about 2 nanoseconds per packet (which is in about the region of a processor cache access).

    Even more problematic however is their set of context assumptions. For example today 100 Gb/s is not "big" for a core router. There are quite some routers with multiple 200 Gb/s line cards.

    If you really knew about throughput optimization you wouldn't have selected and proposed (as "wisdom") that paper. Have a look at their algorithm again! Multi-level loops which boil down to trashed caches - one of the things to urgently avoid. I stopped reading there.

    In case you are technically interested: No matter how one tries that game the decisive issue is to avoid memory reads and writes and to do as much in the caches as any possible. That translates to two approaches both of which can be used together, (a) fast (albeit expensive) memory (SRAM instead DRAM), and (b) specialized processors (highly likely multiple FPGAs) as routing is largely a lookup operation (as far as processing time is concerned).

    As for academic papers I suggest to be careful, mainly for two reasons. For one, academia is regrettably focussed on publishing, publishing, publishing, which leads to many, uhm, not exactly significant papers. The other reason is that the academic approach is quite different from the industry approach and often some approach that looks great in academia has little worth in the industry (e.g. for practical reasons).

    And IPv6 anyway very much increases workload and complexity. I'll provide an example that is directly related to the paper that impressed you: 128 bit tries are way too (compute) intense and also the well understood hashing algorithms are inefficient (for the sake of fairness: the latter is also true for IP4).

    Btw, what for all that trouble? We've been told that the IP4 scarcity will kill the internet very soon since far more than a decade, yet we are still writing here and at least in my case using an IP4.

  • user54321user54321 Member
    edited August 2019

    @jsg said:
    Btw, what for all that trouble? We've been told that the IP4 scarcity will kill the internet very soon since far more than a decade, yet we are still writing here and at least in my case using an IP4.

    I don't know, you can't buy gear at that level where it would matter without IPv6 support anyways so it doesn't matter how complex it is, so even if YOU do IPv4 only in 20 years, nobody else will and there will be no alternative to IPv6 come because it would have the same problems IPv6 had 20 years ago, only that nobody needs a alternative for IPv6.

  • jsgjsg Member, Resident Benchmarker

    @user54321 said:
    I don't know, you can't buy gear at that level where it would matter without IPv6 support anyways so it doesn't matter how complex it is, so even if YOU do IPv4 only in 20 years, nobody else will and there will be no alternative to IPv6 come because it would have the same problems IPv6 had 20 years ago, only that nobody needs a alternative for IPv6.

    Maybe. Or maybe IPv6 will never really take off. We'll see.

  • IPv6 is good for ColoCrossing

  • rsync_dot_netrsync_dot_net Member, Host Rep

    We deployed IPV6 in 2012 or so ... first at our Fremont and Denver and Hong Kong locations which was very, very easy because our provider is he.net and ipv6 is a cause they are championing.

    Then a bit later in our Zurich location which is served by init7 bandwidth.

    Very, very easy integration because we do not own or operate any network equipment - all we have are dumb, unmanaged, top of rack switches.

    All these years later, I think there are maybe 10 or 20 customers globally that connect to us over ipv6. Pretty sure they just do it for the novelty.

Sign In or Register to comment.