After reading this I think any character I make will have an implanted commlink with no wireless and use a datajack or trode net to DNI the implanted commlink into my PAN's commlink on an as-required-for-that-moment basis.

Pretty advanced matrix knowledge, Rotbart! (IMG:style_emoticons/default/smile.gif)

Don't.

Datajacks are bad.
Per Unwired, a Datajack acts as a direct(!) shortcut to every implant with DNI. Thus ignoring Clustering, Slaving and worse, Layered Access... even if they have no wireless interface at all.

Get Skinlink on the internal commlink and use that to connect to the outside world in addition to it's wireless interface for emergencies... all other implants should have no wifi at all. If the fit hits the shan, you can turn off skinlink and wifi completely, leaving any would-be intruder out.

Then use the commlink as chokepoint for a layered access topology, leading to the cluster of your senseware, which susequently leads to the cluster of lifesustaining implants like biomonitors, autoinjectors and nanohives. Out that node would lead the way to the real important things, like skillwires (running Activesoft) and the sim module (running the Biofeedback filter and Know-/Linguasofts) - of course, those node are normally not accepting any connections at all, since there is no need to in AR. (Always remember, once a hacker has reached your sim module, he can force you into VR and club you to death with Black Hammer - Cold ASIST only slows that down as the cold Stun Damage will overflow as usual... ony Blackout doesn't.)

Of course, every implant should have admin access hardwired to DNI in addition to the basic DNI triggers and every node should be 24h strong encrypted, running Ergonomic Analyse, Encrypt and Stealth (to obscure the layout) at least. The sensware cluster will run all the Sensor Softs and the TacSoft, as well as some ICs - and so should every other node. Those ICs then can access the Commlink to defend it.

Looking at unwired... it seems the only advantage of slaving is allowing you to protect lowsec peripherals, by creating a dedicated firewall device then slaving all the external weak links through it. I can't see anything in it which increases your system limit or the like.

Quite frankly, I would completely remove cyberwares signal chips... there's zero reason to have even signal 0 on most of them. (yeah can we say I'm not a fan of wireless everything and anything)

But, clustering does look a little interesting. Just for sake of simplicity I'll refer to sub-nodes of a cluster as blades.
For a standard node... Processor Limit == Response.

Looking at clustering... okay.
Response: average of devices
System & Firewall: minimum of the devices
Process Limit: (Sum of Blades)/2

So now, just for arguments sake I'm trying to build a beowulf cluster out of commlinks.
Response5 (R5), System5 (S5), Firewall6 (F6)

1st: R5, S5, F6, Proc5, Per1
2nd: R5, S5, F6, Proc5(10/2), Per2
3rd: R5, S5, F6, Proc8(15/2), Per3
4th: R5, S5, F6, Proc10(20/2), Per4
5th: R5, S5, F6, Proc13(25/2), Per5
6th: R5, S5, F6, Proc15(30/2), Per6

I'm not quite clear how running multiple personas on a single system is of much use to the typical decker, please correct me if this is wrong or what I'm missing. So this strikes me as not a very usefull reason to cluster. Also unwired itself states, that a standard node can only run one persona at a time, so unless this can now be treated as a special kind of nexi it's pointless (an actual single device nexi doesn't have system limited by response, and some other bits while the cluster of nodes has these determined by worst spec). It's also good to note that it's more expensive than buying a nexi.

So just for arguments sake... using numbers above. What would happen if a decker would try to setup 5 commlinks as a cluster node? Say they're packed into his cyberarm (one of the few reasons I can see to do this). Is it now treated as a nexi (R5, S5, F6, Proc13, Pers5) as seems reasonable.

The biggest thing I can see... is how does the enhanced Proc Limit interact w/ system... normally system limits available processes... but if it changes to proclimit, I can see some use for it.

That's the point of clustering to a single user - with 3+ Devices, the Processor Limit increases.

And given the Ergonomic Option, that can mean a lot.

Read the text on page 102... the datajack only acts as a means to communicate w/ the cyberware. It never once states what 'immediate access is'. It doesn't say that it allows you to immediately assume full and utter control, only that it provides an immediate access route to the device.

I'm waiting for an errata & clarification on this as this strikes me as an unreasonable reading of this entire paragraph.

You are effectively arguing that simply plugging into a datajack gives immediate and full admin access to a cybercommlink. That means it is impossible for any decker to ever use a datajack to directly access any secured system and hack it with said gear, because said system has full access rights to his gear and the first script simply accesses his system at full admin and kills it. This reading effectively also countermands every previous understanding of the relation of cyberware and datajacks in every previous edition when no text present in the sentence says anything explicitly to this effect.

Also as someone who actually knows how to diagram a sentence and a paragraph in english (lost art I know). You're reading too much into things. It says it provides an immediate means of access. It doesn't say it provides no holds barred admin access. I'll also point out sentence 4 does not bar DNI cyberware from being slaved. Just slaving adds a special rule which states that you can only directly hack a slaved device at +2 difficulty with direct access (which datajack/dni would provide). Providing you a means to bypass the slaving but the slaving would still make it harder.

The main advantage of slaving is that it forces an attacker to come through your chokepoint. Very nice. Implants can simply go without wireless.

The main advantage of clustering is that you have a single node that a) can be defended as one entity, and b) can run a large number of programs / personae from one node. The former is interesting for deckers, because the IC does not need to run analyse on multiple nodes. The later lets you get some use out of low-rated nodes, which could never defend themselves on their own anyway.

Now make your own life simple and assume that any hacker that defeated your chokepoint will easily defeat all minor nodes, no matter what you do.

Ryu: that makes a lot of sense... If you have multiple nodes to defend.

Though I might go one step farther and have one isolated note that can only be accessed through the primary node where you keep your sensitive storage. Pretty much the basics of which go like this... when the alarms go off. sensitive storage node grabs the access log, and then immediately cuts off all external links, goes autistic, and shuts down.

I just thought of another 'fun' use of clustering... "okay boss he's clean, he's not carrying anything but some credsticks. " Little do they know the credsticks are actually a system6, firewall6 functional standard node. I think the rules are fairly clear that clustering peripherals only results in a standard node (not a nexi).

But yeah, makes sense... cluster your good stuff which can survive the weakest link/averaging problems. Then slave the rest. Use the multiple personas to increase the subscription limit as necessary (can link systemx2 per persona). I'm still trying to make heads and tails of this though, guess I need a chance to actually try this in play and not just theory craft right now.

I was looking at this from the perspective of... can I form an ad hoc nexi as needed. Say out of all of the teams commlinks on the run. Or build a dedicated mini-beowulf cluster hidden in a cyberlimb. (strikes me that actual nexi either require being carried/placed in a dedicated drone which can be targetted and destroyed... or as a worn computer like a laptop which is also 'obvious' and a target for attention.

No one is saying that a datajack cannot be defended; he's just saying that once it's beaten, the consequences are pretty dire. I personally think the "Immediate access" rule was designed so that cyber-hacking would become viable, since nearly everyone has their wireless turned off.

I'll also add that I think a datajack is redundant when you have an implanted commlink. With a sim module, you can run all 'softs on a trode net, or a commlink. That's really the only advantage to a datajack, and it's not a very big one.

So let me get this straight... you don't need the datajack to plug into secured systems then as a decker?

I'm guessing you're going from your internal commlink... relaying to a slaved external one and plugging the fiber into that one and then into the target.

If that's the case, yes it is pointless. I guess I have the old image of the decker plugging himself directly into secured systems to break into them too firmly entrenched into my mind.

Clusters are definitly for those with many cyberware implants. You are after all halving your processor limit, on top off the even more noticeable drop in the number of "slots" for ergonomic programs. The dedicated hacker/rigger will run a cluster, while someone with only eyes/ears/internal comlink will not.

Note that having multiple agents will come in handy when you are trying to take over a peer-to-peer structured PAN. You don´t need 16+ dice to exploit a rating 3 device.

While I understand how clusters work in the rules, I don't understand why you'd ever want to create one with your implants and gear. I mean, many of those devices have piss-poor stats and thus the more you cluster together, the worse the node becomes. So what's the point in being able to run dozens of Rating 1-2 programs? How does it improve security of the cluster? Does a dozen Rating 1 or 2 Analyze or Browse programs really make it easier to find something more than, say, a single Rating 5 or 6 program would?

Keep in mind the Matrix rules are my weakest area in the game, so I'm genuinely looking for help in understanding this. (IMG:style_emoticons/default/biggrin.gif)

Additionally, why is it that you can't seem to just pull the plug/rip out the batteries/hit the kill switch on a commlink at the first sign of a hacker? Shouldn't that, logically, completely stop the assault and render you immune from being hacked? Especially if you chose to have the wireless functionality of your implants removed rather than simply turned off.

Lets say your hacker has six cyber-implants of grade alpha (doable). Those can take up one subscription slot on the comlink or six. Now add the cybercomlink, agent command channels, and some spy drones, and you are rapidly approaching your subscription limit.

And alpha-datajackS are a cheap source of rating 4 implants.

Funk... if I understand Ryu's logic. Because it allows you to run one set of defensive utilities to rule them all. I agree with you... I wouldn't want to have to deal w/ all the low system ratings and inevitable weak firewall, and weak response you're going to get out of that node. It wouldn't be able to defend itself adequately IMO. If I understand the clustering rules... for low rating devices (like cyber and peripherals) you're lucky to get one low grade commlink out of a cluster of them.

I'm in the same boat as you, I was waiting for unwired before really taking any time to study SR4's matrix.

As far as I'm concerned... GOOD devices are candidates for clustering. Or the cluster of certified credsticks just for giggles ;P. They can actually manage to have the firewall/defensive program ratings to matter. Anything weaker than that gets slaved to the primary defensive node.

Another bit I'm trying to work out is can I run defensive stuff on one supernode... then have it forward deploy itself to a previous node the attacker needs to bust through to get to the good node.

Also not addressed is the 'cluster of clusters' problem... build a cluster node... then build a larger cluster node out of smaller clusters. Or what happens when a cluster is crashed... (does it fragment into lots of individual devices... or does the system reboot as a whole.... I suspect reboots as a whole due to the skill check required for the advanced configuration in the first place). I similarly don't see any mechanic to degrade the node by selectively crashing it's constituent parts.

You can join clusters by making all devices into one cluster - no other way (You could argue otherwise, but it would only lead to you halving the processor limit twice).

You should look into using two clusters, "high grade devices" and "all others".

Ryu: depending on how the systems fragmented when the cluster goes... it might be preferable to have a smaller shard cluster left defending itself... than just a loose device. Might be worth the double halving of proc limit... Probably not but it's just conjecture.

But seriously... what kind of matrix stats are you going to get out of your "all others" cluster?! How is that better than just slaving them individually to the 'high grade device' cluster node. (remember that cluster node has a good many subscriptions available)

The "all others" cluster is there because I don´t want to bother with the individual devices. That´s why they ended up there.

Now read "slaving" again, and tell me why you don´t slave "all others" to "high grade devices".

(You do - There is something to discover, think about the options of gaining access.)

That's what I'm talking about and as it completely defeats layered access, it bad enough.

You pretty much got it. I tend to think old-school as well, but the reality is that datajacks are redundant.


The advantage of a cluster is the ability to run multiple personas on it. You can create a variant of the hackastack/Agent Smith trick, by loading your persona with a dozen agent personas. Since your persona limit increases rapidly, you can legitimately run a ton of agents on your persona, and agents run on your persona aren't restricted by their access ID.



It's a trivial matter to get a hold of a Firewall 6, crack the copy protection, and then upload it to every device you own. Since Firewall isn't restricted by System or Response, and doesn't take up program slots, it's actually a nice trick.



There's nothing that says you can't turn a cluster into a chokepoint for another cluster. In fact, it's a smart idea, since your persona limit increases so easily. You can load down a cluster with IC easily.

Okay this is the part I don't think is correct... one I don't buy that you can just willy nilly replace firewall on embedded peripheral nodes or cyberware running firmware as opposed to an operating system. The description says they can only run programs they're designed to and nothing else, and I don't think a peripheral is designed to run a firewall designed for a commlink. You're talking about running a standard program on something individually limited to running firmware (it's not a cluster until after it's an individual device).

I'll point this out... bodyware specifically has a device rating of *1* (p214 BBB, and I don't see anything in the tables in unwired to change that). Headware is a reasonable 3, so I can see forming those into usable clusters.

I don't see the same problem for say, upgrading all your friends commlinks which are standard nodes capable of running GP software in their own right.

I see my error, for peripheral nodes... I was under the impression once clustered they only functioned as a standard node able to run a single persona. Now I need to reread the agent and persona bits, and work out exactly what persona limit means and benefits... as I quite frankly don't understand those well yet. So any tidbits there to watch out for appreciated. But even at that... I think you end up with the matrix equivalent of a mage w/ a pack of watchers... lots of bark very little bite.

I believe the slaving rules are custom made for protecting the weakest of your links. Clustering them is supposed to be a measure of last resort in the absence of a commlink or other computing device. This sort of follows onto the next one.



The problem here is the low firewall, system stats. See above, I don't think you can just upgrade them to system6/firewall6 all that easily or cheaply. They don't follow the system is limited response rule because they run specialized firmware. I don't think you can abuse that rule to suddenly upgrade them all. Even if you could get a good firewall you're limited to an analyze of only 1 or 2. Or very low rating agents.

I've reread the slaving bits... 3 options...
Okay hack the device at +2 threshold... but ONLY IF YOU HAVE WIRED ACCESS (good luck getting wired access to a smartgun, especially if it's skinlinked and the person has intentionally damaged the fiber port, for a lot of peripherals there's zero reason for fiber access so destroy the port or render it unusable IMO). Granted a peripheral node like a smartgun only has an admin account so it doesn't have that nasty +6 threshhold.

Hack the master defense node... which the slave forwards all comms attempts to. Said device should be able to handle these though I'd think.

Spoof commands from the master to the slave, (good luck especially if it's skinlinked so you don't have access to it's command channel, or if it gets it's commands from DNI such as cyberware and you've ripped out the wireless completely (or hit it with EMP/RFid erasers to blow out the transmitters intentionally!).

What am I missing... the hacking the slave DOES NOT give admin access to the master node. The master node has admin access to the slaved device but not vice versa.

The lack of anything except an admin login on a peripheral node, so the lacking of that +6 threshold to gain access is a huge problem IMO. That weakness would extend to a cluster built of them IMO.

Thanks to all for the time, I'm learning a lot.

How is this better than doing the same thing without the cluster? Especially if you buy an Agent with the Adaptive programming option. I mean, if you cluster five Rating 3 and five Rating 1 nodes, you basically wind up with a Rating 2 node that can support half as many personas/programs. Wouldn't it take a hacker longer to completely infiltrate your gear and implants if they have to break through one after another, especially if you also load each one with Firewall 6?

I mean, it just seems like it'd be wiser to make a "chain" of slaved nodes. Your commlink being the true master, with your next highest rated node slaved to your commlink, the next slaved to that one, and so on and so forth. Wouldn't that definitely make them have to fight through everything? Especially if you bury your most dangerous implants (datajacks, cyberlimbs, smartlinks, cybereyes, etc.) at the end of the chain?

The idea is that you slave the weak cluster to the strong cluster. The strong cluster has admin access anyway, either by cluster or individually.

Clustering makes you loose insignificant processing power (you have enough with comlink + better devices), and reduces your bookkeeping. Anyone who attacks the weak node already went through your comlink (only device with matrix access), and the controlling stronger cluster. At that point the battle is simply over.

One other use for clustering is communications. Don´t use the comlinks for wireless communications, but dedicated com units. Cluster them, and have the hacker guard the combined node. Way safer than individual connections.

The special four are, well, special. You can run a System 6 program on anything you like; it's just that the Response cap will prevent you from using it fully. You can run a Firewall, however, at full effectiveness. Every device has a rating that stands in for Firewall, but there's no reason why you can't use the same Firewall program on a commlink-like device. In other words, anything with a Device rating is fair game for uploading a Firewall 6.

You use it as the chokepoint, but not the only one. Clustering allows you to also go on the offensive, with an Agent Smith army.


Actually, it's better to use Rating 1 commlinks for the links in a chain. The trick is, you can only subscribe to 2 things at a time, and you use up those slots when it links to the rest of the chain. In order for someone to hack you, they need to subscribe to you first; in the process of doing that, they change the list and immediately inform you that something's going on, as the rest of the chain "drops off".

.


I was curious about this and ran across in Unwired, p55 where it says if you exceed your allowed # of subscriptions additional subscriptions count as programs:

Cain: My POV is simple. Peripheral nodes require special systems software and I think that Unwired says as much on page 48... "They are only able to run a single persona, and can ONLY run programs they are designed to use" also the bit stating that system and response are unlinked explicitly states it's because they are so very limited and specialized. But you're telling me that system and firewall software designed for an 'unlimited' commlink is no problemo.

To me peripheral nodes includes bodyware and headware. Ones only device rating 1 and the other 3. I have nothing against upgrading the firewall and system to 6... but if I was GM'ing I'd say you'd need to hack an upgraded bodyware/headware firewall. And I wouldn't feel that I was using GM fiat at all, but that the rules had me fully covered. Or simply, don't be cheap and just pay the 3000 right away for the rating 6 firewall upgrade when you buy the cyber in the first place. (3000 for upgraded firewall6 and another 600 for say system3 upgrade on bodyware firmware when the cybers worth say 40k to start with isn't asking all that much).

I'd also say this cuts both ways... someone who does hack cybereyes is severely limited in what they can do. They can use their edit program to brute force blind the guy... or more advanced users could selectively edit the field of view unobtrusively (such as making items disappear, displaced, or craft illusions). Use the device as extra storage... wipe it... etc. I'd balk at them trying to run an agent or hide a tracer there.


Also quite seriously, either the limitations for peripheral nodes have meaning or they should have never bothered printing that sentence. As simply clustering two means you can now run anything you damn well please which doesn't strike me as something the authors ever intended. It strikes me that running IC and advanced intrusion software probably isn't an option for peripheral nodes.

Case in point: you have cybereyes and cyberears... both are device rating 3. They can only run video and audio specific software respectively. Now just because you cluster them... you can run anything your heart desires?... I don't think so. Strictly reading RAW they can't even run an agent or security outside maybe encrypt (full node encryption) as it's not related to optical processing/image enhancement. (the only thing the cybereye is designed and optomized for).
I similarly don't buy that you can load generic firewall Mk6 onto the cybereyes or ears because they run special OS firmware, not generic commlink computer OS.

There's another issue here which isn't addressed... but clusters are subject to the weakest security link rule. It's never stated, but once you start incorporating peripheral nodes into a cluster, I don't think you can create anything except admin level accounts and lose the no +6 difficulty to hack penalty.

I'm asking this as an open question... what are the limits to software that can be run on a peripheral node. And how do those limitations transfer once clustered? I'm truly curious how you would make those limitations worth the ink they're printed with. (call it FAQ material eventually as I think it's pretty unclear, probably intentionally vague to give GM's wiggle room, I think it's pretty gray myself).

Or maybe I'm just being overly nitpicky... it's only 300 to obtain the warez rating 6 firewall in the first place, and only 50 a month to keep it patched. An extra 150 a month to maintain 3 different firewalls (bodyware, headware, and general use) doesn't strike me as all that burdensome. But it could also be argued, that it's not the one, but the other 10 hacking utilities costing 100 a piece a month to keep at rating 6 which are adding to the costs. (quite frankly I think the SOTA works quite well and they did a bang up job of keeping the costs reasonable... 10 hacking utilities for 1000/month extra in monthly living costs. If not paid they only degrade 1 rank and you can pay extra to bring them up when you have the scratch, is pretty easy for a hacker... especially one spoofing up his own lifestyle. (really spoofing low to middle isn't that hard, you're still saving 2000 a month even after software over just buying mid!)

Yeah, I saw that one, too. On a Rating 1 commlink, that means you just went over your program limit, and Response just dropped to 0. For all intents and purposes, that 'link just died, and it's blatantly obvious to you.


Technically, yes. They're not the same as Pilot programs, which are specific to a given vehicle class. But otherwise, a program is a program is a program.

I find that interesting, because I remember a discussion about running IC on your cyber as a defense against intruders. You don't need to cluster it first. IC is effectively the same thing as an agent, so if you'd allow one, why wouldn't you allow both?

In theory, you can run any program on any computer. Other than very specific restrictions (like running an agent on a vehicle), you can run any program on any Device. In practice, most Devices are going to be so weak, running anything on them is probably futile. (Although I do wonder about this-- if most Devices have a rating of 3, doesn't that mean they're better than low-end commlinks? And cheaper, too?) The persona limit indicates that you can run a persona on peripheral nodes, so putting an agent on it should be fine.